Laminated glass : dynamic rupture of adhesion

Paul Elzière

To cite this version:

Paul Elzière. Laminated glass : dynamic rupture of adhesion. Polymers. Université Pierre et Marie
Curie - Paris VI, 2016. English. �NNT : 2016PA066310�. �tel-01471672�

HAL Id: tel-01471672

https://tel.archives-ouvertes.fr/tel-01471672
Submitted on 20 Feb 2017

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci-
entific research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
               

Université Pierre et Marie Curie

École Doctorale 397 : Physique et Chimie des Matériaux
Saint-Gobain / Laboratoire SIMM

Verre feuilleté :
rupture dynamique d’adhésion

Paul Elzière
Thèse de doctorat en physico-chimie des polymères
Dirigée par Matteo Ciccotti, Étienne Barthel et Cécile Dalle-Ferrier

Présentée et soutenue publiquement le 29 septembre 2016

Devant un jury composé de:

R. Estevez Professeur –INP Grenoble Président du Jury

L. Leger Professeur émerite –Université Paris Sud

Rapporteurs
T. Pardoen Professeur –Université Catholique de Louvain

C. Maurini Maître de Conférences –IJLRA/UPMC Examinateur

M. Ciccotti Professeur –ESPCI Paris Directeur de thèse

E. Barthel Directeur de Recherches CNRS –SIMM/ESPCI Paris

Co-directeurs de thèse
C. Dalle-Ferrier Saint-Gobain Glass

This work is licensed under a Creative Commons

Attribution-Non Commercial 4.0 International License
Laminated glass: dynamic rupture of adhesion

Paul Elzière
Matteo Ciccotti, Étienne Barthel and Cécile Dalle-Ferrier

September 29th, 2016

Remerciements
Je voudrais remercier tout d’abord mes trois encadrants Matteo Ciccotti, Étienne
Barthel et Cécile Dalle-Ferrier.
Je souhaite témoigner ma reconnaissance à Matteo Ciccoti, pour sa confiance,
ses conseils précieux, son enseignement patient de la mécanique des polymères et
de la fracture, sa bonne humeur constante et la passion communicative dont il fait
preuve dans son travail.
Je veux faire part de ma plus profonde gratitude à Étienne Barthel pour sa
disponibilité, ses innombrables conseils, son soutien sans faille et surtout sa patience
tout au long de ce projet lorsque, trop souvent, je venais interrompre son travail.
J’aimerais adresser à Cécile Dalle-Ferrier mes plus chaleureux remerciements
pour la confiance qu’elle a bien voulu m’accorder, pour avoir initié, poussé et ac-
compagné ce projet, pour la chance qu’elle m’a donnée de travailler au plus près de
l’outil industriel ainsi que pour les discussions que nous avons eu sur le monde de
l’entreprise et tant d’autre sujets.

Je voudrais remercier ensuite les membres du jury qui ont généreusement pris
le temps de relire et de commenter ce document ainsi que pour les discussions que
nous avons eu lors de la soutenance.

Ce travail n’aurait pu avoir lieu sans le soutien financier de Saint-Gobain Recherche.

Je souhaite remercier Mathieu Joanicot et François Creuzet, directeurs de la recherche
du centre de R&D d’Aubervilliers, pour leur soutien et leur suivi du projet.
Je tiens à remercier Vincent Rachet pour avoir initié cette thèse et avoir soutenu
et suivi le projet mais aussi pour ses conseils sur mon projet professionnel et son
accueil chaleureux au centre R&D de Sully-sur-Loire.
J’ai eu la chance de pouvoir intégrer une équipe industrielle qui m’a toujours
accueilli avec bonne humeur et je tiens à remercier toute l’équipe PIL de Saint-
Gobain Recherche et en particulier : Joel Azevedo, Julien Beaumont, Joana Girard,
Hans Herbert, Diamante Mace, Caroline Parneix, Joel Robineau, Cyrielle Rudaz,
Tamar Saison, et Leila Tahroucht. Je voudrais aussi remercier tous ceux qui à Saint-
Gobain m’ont aidé pour mes expériences, pour la compréhension et la modélisation
des phénomènes observés ou avec qui j’ai eu le plaisir de discuter : Alessandro
Benedetto, Ivan Berline, Pierrick Cavalie, Samuel Dubrenat, Alexandre Kerambloch,
Jean-Yvon Faou, Maxime Van Landeghem. . .
En particulier, je veux remercier Nathalie Dideron qui a eu la patience de me
former à mon arrivée et qui, avec gentillesse, a su m’encourager et m’aider durant

i
ces trois années et sans qui ce projet n’existerait pas.
Je souhaite aussi remercier Marie Lamblet et Hélène Lannibois-Dréan pour l’intérêt
qu’elles ont porté à ce projet, leur suivi assidu, leur aide et leurs conseils tant sci-
entifiques que sur le monde de l’entreprise.
Je voudrais remercier Alexis Chenneviere et Keyvan Piroird pour leur aide et
leur contribution à mon travail. Je veux aussi exprimer ma gratitude à René Gy,
pour son intérêt pour le projet, pour avoir participé à tous mes points d’avancements
et ses remarques et conseils. Je voudrais également adresser mes remerciements à
Romain Decourcelle qui a lui aussi travaillé sur ce sujet durant sa thèse, pour son
suivi de ce projet avec intérêt et les échanges que nous avons pu avoir. Je remercie
également Jean Charles-Sauvesty pour son aide sur les éléments finis.

Au cours de ma scolarité à l’ESPCI j’ai eu la chance de pouvoir travailler à de

nombreuses reprises au sein du laboratoire SIMM : aussi, je souhaite remercier très
chaleureusement Christian Fretigny directeur du laboratoire et Guylaine Ducouret
directrice adjointe, pour m’avoir accueilli dans cette unité durant toutes ces années.
Mes très sincères remerciements à Costantino Creton pour m’avoir accueilli au sein
de son équipe et pour les nombreuses discussions et ses conseils pertinents ainsi que
pour avoir partagé sa passion du vin et de la table avec nous.
J’exprime ma profonde reconnaissance à Alba Marcellan qui m’a donné le gout
des polymères et de la mécanique des matériaux et qui m’a permis de vivre des
aventures scientifiques riches et passionnantes.
Je tiens également à remercier tous les permanents du laboratoire et en particulier
ceux qui ont toujours eu un moment pour discuter avec moi d’un problème pratique
ou plus philosophique. J’ai pu en effet apprécier au cours de ces années dans le lab-
oratoire, la disponibilité des chercheurs pour m’aider à surmonter les obstacles que
nous pouvions rencontrer. Ainsi je tiens à remercier: Sabine Cantournet, Antoine
Chateauminois, Dominique Hourdet, François Lequeux, Hélène Montès, Laurence
Talini, Yvette Tran et Emilie Verneuil.
Je voudrais aussi remercier les personnes qui font vivre le laboratoire au quotidien
et sans qui nos efforts ne pourraient aboutir: Pierre Christine, Gilles Garnaud, Pierre
Landais et Flore Lasaone, mais aussi Armand Hakopian, à qui je promets de ne rien
modifier sur les serveurs en partant, Freddy Martin esprit et corps de ce laboratoire,
Bruno Bresson et ceux qu’on ne remercie plus, Ludovic Olanier et David Martina
pour leur aide et leurs conseils dans le montage de mes expériences. Je souhaite
adresser des remerciements très particuliers à Mohamed Hanafi pour m’avoir aidé
dans la réalisation des expériences de DSC et dans l’analyse des résultats.

ii
 Cette thèse s’est faite dans la joie et la bonne humeur, cela n’aurait pas été pos-
sible sans tous les stagiaires, doctorants et post-doctorants qui constituent le cœur
de ce laboratoire. Je voudrais commencer par remercier mes co-bureaux présents
et passés: Francisco Cedano, Jessica Delavoipiere, Richard Villey et Jingwen Zhao
qui ont enduré mes sautes d’humeurs. Je voudrais aussi remercier les différents
bureaux de l’escalier H, Romain Dubourget, Natacha Goutay, Jennifer Macron et
Cécile Mussault, Benjamin Chollet, Robert Gurney, Anne-Charlotte Le Gulluche et
Pierre Millereau, Alice Boursier, Davide Colombo, Pierre Gelineau et Robin Ma-
surel, pour leur patience exceptionnelle lors de mes visites quotidiennes mais aussi
pour les bons moments que nous avons partagés au travail et en dehors. Merci égale-
ment à Quentin Demassieux qui a non seulement été un excellent camarade mais qui
a aussi contributé aux expériences de SAXS. Je voudrais aussi remercier mes amis
de la 128ème promotion avec qui j’ai partagé ces trois années au SIMM, Pascaline
Hayoun, Marine Protat et Marc Yonger. Sans pouvoir les citer tous je voudrais re-
mercier les personnes avec qui j’ai partagé rires, bons moments et conseils précieux:
Mélanie Arangalage, Charles Barrand, Laure Bluteau, Xavier Caliès, Julien Chopin,
Rémi Deleurence, Étienne Ducrot, Guillaume Fisher, Hui Guo, Thitima Limpanich-
pakdee, Éric Lintingre, Ekkachai Martwong, Yannick Nziakou, Séverine Roses, Tom
Saint-Martin, Corentin Tregouet, Pauline Valois, Judith Wollbrett-Blitz,. . .

Je souhaiterait également remercier les professeurs Chung-Yuen Hui et Anand

Jagota pour leurs enseignements et les longues et riches discussions qu’ils ont bien
voulu avoir avec moi. Je veux aussi remercier toutes les personnes qui ont pris le
temps de discuter de mon sujet et qui ont participé ainsi à la construction de ce
projet: José Bico, Guillaume Parry, Suomi Ponce, Claire Prada-Julia, Samuel Raetz
et Benoît Roman.
Durant cette thèse j’ai eu la chance d’encadrer et de travailler avec Louis Deber-
trand et Noëlig Daggorn. Je souhaite les remercier pour leur importante contribution
à mon travail. Je voudrais aussi remercier Raphaëlle Kullis qui a réalisé un travail
exceptionnel au cours de son stage.

Un immense merci à mes très chers amis Adrien, Loic, Nathalie, Nicolas, Samy,
Sarah et Yann qui ont toujours été là pour moi. Un très très gros merci également à
Adrien, Nicolas, Thomas, Ugo et Ève pour les copieux repas que nous avons partagés
et les heures passées sur nos jeux favoris. Enfin un très grand merci à Charlène pour
son oreille attentive et son amitié durant toutes ces années.
Il me faut à ce stade avoir une pensé spéciale pour le raton. . .

iii
 Je ne pourrais jamais assez remercier ma sœur et mes parents qui ont toujours
été à mes côtés.
Enfin, je dédie ce travail à mes grands-parents qui sont pour moi une source
d’inspiration et de courage:

Βασιλική και Σταύρος Παπαγιάννης Simone et Pierre Elzière

iv
Contents

1 Introduction 1
1.1 A brief industrial history of laminated glass . . . . . . . . . . . . . . 1
1.2 Impact on laminated glass . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Standard tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Kinematics of impact on laminated glass . . . . . . . . . . . . 5
1.2.3 Previous works on impact . . . . . . . . . . . . . . . . . . . . 8
1.3 Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Methods 13
2.1 Laminated glass assembly . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Silanization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Mechanical behavior of the interlayer . . . . . . . . . . . . . . . . . . 15
2.3.1 Dynamical Mechanical Analysis (DMA) . . . . . . . . . . . . 15
2.3.2 Rheology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.3 Uniaxial tension . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Different models to describe the constitutive behavior of the interlayer 18
2.4.1 Small strain description: Generalized Maxwell model . . . . . 19
2.4.2 Hyperelasticity: Arruda-Boyce model . . . . . . . . . . . . . . 19
2.5 Delamination experiments on laminated glass . . . . . . . . . . . . . 20
2.5.1 Peel test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.2 Through crack tensile test . . . . . . . . . . . . . . . . . . . . 21
2.6 Optical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6.1 Video acquisition . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6.2 Digital image correlation . . . . . . . . . . . . . . . . . . . . . 23
2.6.3 Photoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7 Differential Scanning Calorimetry . . . . . . . . . . . . . . . . . . . . 26
2.8 X-ray scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

v
3 A complex structure and rheology 29
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Poly(Vinyl Butyral) interlayer . . . . . . . . . . . . . . . . . . . . . . 29
3.2.1 Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.2 Hydroxyl groups . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Rheology of the PVB . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.1 Small strain viscoelasticity . . . . . . . . . . . . . . . . . . . . 33
3.3.2 Large strain uniaxial tension . . . . . . . . . . . . . . . . . . . 38
3.4 Strain induced birefringence . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.1 Influence of strain rate and temperature on birefringence . . . 47
3.4.2 Birefringence during relaxation experiment . . . . . . . . . . . 49
3.4.3 Partial conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.5 Evidence of a second phase . . . . . . . . . . . . . . . . . . . . . . . . 53
3.5.1 An exothermic signal . . . . . . . . . . . . . . . . . . . . . . . 53
3.5.2 Evidence through X-ray scattering . . . . . . . . . . . . . . . 53
3.5.3 A schematic model of the structure . . . . . . . . . . . . . . . 55
3.6 A rheological model: two dissipation mechanisms . . . . . . . . . . . 56

4 Model delamination experiment 61

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Description of a typical Through Crack Tensile Test . . . . . . . . . . 63
4.3 Influence of velocity and temperature on delamination: phase diagram 65
4.3.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.2 Comparison with previous studies . . . . . . . . . . . . . . . . 68
4.4 Distribution of the deformation of the interlayer in the TCT test . . . 69
4.4.1 Deformation zone measured by photoelasticity . . . . . . . . 70
4.4.2 Fast stretching zone measured in DIC . . . . . . . . . . . . . . 71
4.4.3 Dependence of the fast stretching zone length on applied ve-
locity and temperature . . . . . . . . . . . . . . . . . . . . . . 73
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5 Energy dissipation during delamination 77

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Macroscopic work of fracture . . . . . . . . . . . . . . . . . . . . . . . 78
5.3 Impact of the interlayer thickness . . . . . . . . . . . . . . . . . . . . 79
5.4 Different zones of dissipation . . . . . . . . . . . . . . . . . . . . . . . 81
5.5 Modeling the bulk stretching of the interlayer . . . . . . . . . . . . . 83

vi
 5.6 Dissipated energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.7 Influence of the temperature and applied velocity on the dissipation
mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.8 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 86

6 Interface modification – Preliminary results 89

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 Impact of silanization on the interface and on the peel work of fracture 90
6.3 Impact of an interface modification on the TCT test response . . . . 92
6.3.1 Different steady state delamination regimes . . . . . . . . . . 92
6.3.2 Steady state delamination for the lower adhesion . . . . . . . 92
6.3.3 A change in the dissipated energies . . . . . . . . . . . . . . . 95
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

7 Finite element modeling description 101

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Cohesive zone model for the interfacial rupture . . . . . . . . . . . . 102
7.3 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.4 Recovering a steady state delamination . . . . . . . . . . . . . . . . 111
7.4.1 Decohesion processes . . . . . . . . . . . . . . . . . . . . . . . 112
7.4.2 Hydrostatic stress induced by the boundary conditions and
the incompressibility . . . . . . . . . . . . . . . . . . . . . . . 113
7.4.3 Energy flows balance . . . . . . . . . . . . . . . . . . . . . . . 114
7.4.4 Far field measurements . . . . . . . . . . . . . . . . . . . . . . 116
7.5 Two zones of dissipation . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.5.1 The fast stretching zone . . . . . . . . . . . . . . . . . . . . . 118
7.5.2 Near crack process zone . . . . . . . . . . . . . . . . . . . . . 121
7.6 Near crack work of fracture . . . . . . . . . . . . . . . . . . . . . . . 121
7.7 Impact of interlayer relaxation time and work of separation . . . . . 124
7.7.1 Work of separation . . . . . . . . . . . . . . . . . . . . . . . . 126
7.7.2 Viscoelastic relaxation time . . . . . . . . . . . . . . . . . . . 129
7.8 Coupling between the near crack and bulk stretch responses. . . . . . 132
7.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

8 Conclusions and perspectives 137

Résumé en français 143

vii
Appendices 155
A Effect of the thermal treatment during laminated glass preparation
on the mechanical behavior of the interlayer . . . . . . . . . . . . . . 155
B Arruda Boyce Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

Bibliography 158

Abstract 164

viii
Table of Abbreviations
Interlayer Behavior

σT Uniaxial traction true stress

T Uniaxial traction nominal strain
λ Uniaxial traction stretch
µ Shear modulus
E Young modulus
τi Viscoelastic relaxation times
λm Maximal chain extensibility

Through Crack Tensile Test and Finite Element Modeling

h Interlayer thickness
w Laminated glass sample width
δ Applied displacement
δ̇ Applied velocity
a Position of the delamination front relative to the glass crack
xF Distance to the point at the vertical of the delamination front
FTCT Delamination force
0
FTCT Delamination force, steady state value
TCT Delamination nominal strain
0TCT Delamination nominal strain, steady state value
Gm Macroscopic work of fracture
Γbulk Bulk stretching work
Γvisco
bulk Viscoelastic losses part of the bulk stretching work
Γel
bulk Elastic part of the bulk stretching work
Πbulk Volumic density of the bulk stretching work
Γcrack Near crack work of fracture
Γs Separation work (molecular scale at the interface)

ix
x
Chapter 1

Introduction

1.1 A brief industrial history of laminated glass

Glass has been used in housings for about 2000 years and in vehicles for more than a
century for its transparency and elegance. However, its poor mechanical resistance
and other issues, in relation to its poor thermal and sound insulation, prevented it to
be used as a structural material in buildings. It is only thanks to some improvements
of these properties that glass was made suitable for use as one of the main building
materials. For example, multiple glazing and the use of an inert gas in between the
two glazings reduce heat transfer. Glass surface treatment enables new properties
such as hydrophilic self-cleaning glasses or active glasses which can be tinted at will
under the action of an electric field. Finally, the mechanical properties of glass and
especially its resistance to impact, were improved through toughened or laminated
glass. It made it suitable for safety applications in housings and vehicles.
Laminated glass and toughened glass are often mistaken even if these two mate-
rials are really different and do not bring the same safety features. Toughened glass
results from a rapid cooling of glass plates which traps tension stresses in the glass
volume and compression stresses at the surface of the glass. This surface precom-
pression increases the resistance to fracture of the material. It can also be obtained
thanks to chemical treatments. Toughened glass is also used because it breaks into
small, smooth and thus less dangerous pieces. However, due to the negligible dissipa-
tion these glass are not suitable for impact resistance. On the other hand, laminated
glass is made of an elastomeric interlayer sandwiched in between two glass plates.
Dissipation mechanisms increase the performance to impact of the glass structure.
The adhesion between the glass and the interlayer furthermore prevents flying glass
parts, which might lead to severe injuries when a glass wall or a windshield is im-

1
pacted. Finally, laminated glass will keep the structural integrity of the glass plate,
which is of special interest in some harsh conditions such as hurricane.

Laminated glass was discovered by Edouard Benedictus a French artist (1878-

1930 ) who dropped a glass flask containing a cellulose based glue. The flask pre-
sented cracks in the glass but the whole structure remained. This discovery occurred
in 1903 [1]. Several patents were written beginning in 1909 and the Société du Verre
Triplex (a company partially bought by Saint-Gobain in 1928) started to commer-
cialize the Triplex glass for car windshields. Back then, laminated glass was made
of two glass plates of 2.5 mm thickness and the glass surface was cleaned and coated
with a 10 µm gelatin film which was dried before another wet collodon based layer
was applied. This last layer was used as an adhesion primer. Then a celluloid layer
was put in between the two glasses. The laminated structure was finally immersed
in alcohol which helped remove air bubbles before it was pressed while heating. A
mastic was also applied on the edge of the glass to protect the interlayer from ex-
ternal atmosphere. The celluloid interlayer presented at least two drawbacks. First
it yellowed while aging and it was also delaminating from the glass quite easily. It
was replaced by a Poly(VinylButyral) (PVB) in the 40’s-50’s.

This is still the mostly used interlayer for laminated glass. Other interlayers such
as Polyurethane (PU) or EthyVinylAlcool (EVA) are used for different applications.
Indeed, laminated glasses are used in a wide range of applications. In cars, it is
used for windshields, for roofs and in some lateral car glass on premium cars. In
housing the variety of applications is also quite large. Laminated glass is used for
the external glass of buildings as well as inside panels and walls. Its high resistance
to static loading and impact makes it suitable for floors and ceilings. Polyurethane
is used for plane windshields or other high performance applications. EVA is mainly
used in solar photovoltaïc panels and for lower impact performance and lower cost
applications. Printed interlayers are also available and thus laminated glass can
be used as a decorative piece inside the housing. Some famous example of usage
of Saint-Gobain laminated glass in remarkable buildings are the Louvre Museum
pyramid, or at the first floor of the “Tour-Eiffel”,...

It can be seen through this brief history of laminated glass that the industrial
"know how" is really old. However the mechanical response of the laminated glass
structure during an impact remains a puzzling question. Impact is a complex prob-
lem which involves different mechanical processes: glass rupture, interface breakage
and interlayer stretching.

2
1.2 Impact on laminated glass

1.2.1 Standard tests

Because laminated glass is mainly used as a safety product, it has to comply with
different standard tests depending on the final application. These tests may slightly
vary from one country to another. Their general aim however is hopefully to repro-
duce the type of loading to which the laminated structure will be subjected during
an impact. We will present here two different tests used in France and Europe to
qualify building glasses.

Pendulum test – EN 12600 [2]

The pendulum test is supposed to reproduce the impact of a human body falling
onto a window or a glass wall. The functionality that is tested aims at preventing
the fall of a person through the glass and the injuries which might be induced by
the glass splinters.
The tests are performed on samples of approximately two square meters. The
tests are conducted at room temperature and they have to be repeated at least on
four samples of the same laminated glass. A weight of 50 kg with two inflated tires
around it (which are supposed to reproduce the mechanical behavior of a human
body), is dropped from an increasing height on the glass (Figure 1.1). Three falling
heights are considered and the materials are ranked according to the maximal falling
height they can sustain. It does not necessarily mean that the glass is not broken
but that if it does, it will break in a specific manner which hopefully will not arm
the occupants of the building.

Ball drop test – EN 356 [3]

This test is used to qualify the resistance of laminated glass against manual attack.
Thus its main target is the use of laminated glass in applications such as showcases or
ground floor housings. A steel ball of 4 kg and 100 mm diameter falls from different
heights on an approximately one square meter laminated glass plate (Figure 1.2).
The ball is dropped three times on the glass at the summits of an equilateral triangle.
The test is conducted at ambient temperature. The test is passed if the ball does
not go through the glass in the five seconds following the impact. The laminated
structure is ranked from P1A to P4A which corresponds to drop height going from
1.5 m to 9 m (A P5A class is also used for structure holding 9 drops from 9 m).

3
Figure 1.1: The 50 kg pendulum test used in European Standard EN 12600. A weight
surrounded by two inflated tires is dropped from different heights on a laminated
glass plate [2].

Figure 1.2: A 4 kg ball test used in European Standard EN 356. The ball is dropped
three times on the glass which has to sustain the impacts [3].

4
 A similar drop test is done with laminated glass used for windshields (ECE R43)
in which a smaller steel ball modeling a small rock hitting the windshield is dropped
from different heights onto a smaller piece of laminated glass. The glass has also
to prevent the penetration of the ball but also it must show limited shattering or
glass. This test is performed at -20 ◦C, 20 ◦C and 50 ◦C as the windshield might be
subjected to various temperature conditions.

It can be seen that even if a lot of care is put into these tests in terms of
controlled environments and testing procedures, these tests remain qualitative. In
case the glass fails to pass one of these tests, it is difficult to identify the origin of the
problem. Moreover, it is difficult to predict the behavior of a glass before it is tested
in these specific conditions. This means that in order to develop new products a lot
of trials in large dimensions have to be made.

1.2.2 Kinematics of impact on laminated glass

From these tests, a first picture of what is happening during an impact can be drawn.
During the ball drop standard test for example, the highly complex process of the
impact can be divided into different stages as described in Figure 1.3. Three main
steps might be considered:

0s 10 µs 100 µs 10 s

Figure 1.3: Schematic view a ball drop test impact and corresponding high speed
camera frames.

• In a first step the impacting object hits the glass. This step lasts for a few tens
of microseconds during which the cracks propagate towards the sample edges.
The crack propagation velocity might be affected by the interlayer but it will

5
 be of the order of magnitude of the Raleigh wave speed in glass (5000 m s−1 ).
During this step the whole plate can be considered as rigid as it does not show
yet a large bending.

• In a second time (>100 µs), the glass plate undergoes a large bending. The
plate reaches a maximal bending of about 10 cm after 1 s. Broken parts of glass
are still holding on the interlayer and are pulled apart from each other due to
the plate bending. Meanwhile, the interlayer delaminates from the glass and
is stretched in between the glass parts.

• In the final stage of the impact, the main questions are the following: is the
impacting object going through the glass? Is the object bouncing back the
glass? Are large parts of glass detaching from the glass? Does the glass pass
the standard test? These are questions of industrial relevance.

Basic understanding

Optimal
Impact Performance

Pe
g
rin

rf
te

or
at

ati
Sh

on

Adhesion

Figure 1.4: A balance in the interlayer/glass adhesion is required to reach good

impact performances.

A first understanding of the problem arises from the facts that the impacting
object should not perforate the glass and that no large glass part should detach
from the structure. As underlined by [4], the adhesion between the glass and the
polymeric interlayer must be balanced (Figure 1.4). Indeed, if the glass and the
interlayer are too tightly bonded together the impacting object will go through the
structure as if there was no interlayer. On the contrary, if adhesion is too low, the
broken glass parts will be ejected from the structure. There is a generally admitted
understanding that the impact energy is not dissipated, in these both extreme cases
because the interlayer can not be stretched in both situations.

6
First results: order of magnitude

Beyond this very basic understanding of the problem, several works have already
focused on the resistance of laminated glass to impact. It is difficult, during a real
impact or a standard impact test, to implement measurements of plate deflection
or of interlayer delamination, because of the speed of the impacting object (about
10 m s−1 ) and because of the complex and not predictable crack pattern. That is
why smaller impact tests or bending experiments have been implemented to give a
feeling of the processes at work during rupture. We can draw from these simpler
experiments some orders of magnitude for the delamination of the interlayer and
bending of the glass plate.

Delamination length: On a laminated glass in which a single crack has been

propagated into the glass on each side, a small delamination of the interlayer can
be observed on each part of the crack even if there was little bending (about 1 mm)
applied on the glass to propagate the cracks (Figure 1.5). The delamination length
in this case is about 0.2 mm.

0.2 mm

1 mm
Glass crack

Delamination front

Figure 1.5: Delamination front and glass crack after propagation by undergoing a
1 mm bending. This image was taken under a microscope on a broken laminated
glass sample. Here, the focus is made at one of the PVB glass interfaces close to
the crack that has been propagated in the glass.

Small impact test: Information can also be derived from small impact tests such
as order of magnitude of the time and space scales involved during the impact. In
this experiment square plates of laminated glass of about 10 cm width were used.

7
 The broken laminated glass plate can be considered as a two piece laminated glass
bridged by a stretched interlayer as in Figure 1.6. The plate maximal bending is
about 10 mm. If 200 µm of interlayer is delaminated, then the interlayer is stretched
up to 500 µm which corresponds to a strain of 150%. From this it appears that large
strain behavior of the interlayer has to be considered in this problem. The impact
speed is close to 10 mm s−1 and thus the traction speed on the interlayer is about
500 mm s−1 . Similar order of magnitudes can be derived from what is happening
during a standard ball drop test.

Stretch Length = 0.5 mm

Bending = 10 mm
Plate size = 10 cm

Figure 1.6: Schematic view of a bent and broken laminated glass.

1.2.3 Previous works on impact

More in depth studies on the impact resistance of laminated glass can be found in
the literature.

The laminated glass is often considered as a sandwich object in which the inter-
layer/glass interface is supposed to be perfect. The laminated glass structure is thus
considered as a sandwiched of two perfectly elastic plates with a viscoelastic layer in
between and delamination is not taken into account. This approach is appropriate
to model bending before rupture (large glass thickness or small impacting object
velocities) or for constant loads (wind, snow, . . . ).
In his PhD conducted with Saint-Gobain, Vidal [5], has developed an analytical
and a finite element model of the impact, based on beam and plate mechanics.
The interlayer was described as a linear viscoelastic material. The purpose of this
model is to predict glass rupture. Vidal was able to predict the experimental impact
behavior of the laminated plate with both hard and soft impacting objects, before
the glass rupture. The whole stress field, calculated by this finite element model,
can then be used to predict the glass rupture. A statistical model is used to describe
the rupture probability based on defects present in the glass.
Flocker has developed a similar description of the impact. In [6], he used a lin-
ear viscoelastic material to describe the interlayer and described the glass/interlayer

8
interface as perfect. The effect of different parameters on the laminated glass resis-
tance to impact were studied: impacting object size, interlayer properties, thickness
and boundary conditions. In particular, this work focused on the prediction of the
maximal principal stresses.
Hooper has developed a finite element model of laminated glass using shell ele-
ments [7]. In a first step, the structure was modeled as a perfect laminate in order to
study glass rupture. In a second step however, the stiffness of the broken glass ele-
ments was locally reduced to zero and a rate dependent plastic material was used to
describe the interlayer. This work emphasized the fact that large dissipation occurs
as soon as the interlayer delaminates from the glass and is stretched under impact
loading. However, Hooper does not study the relationship between the mechanical
behavior of the interlayer and the glass/polymer adhesion.

A different approach was considered by Nourry [8] and Decourcelle [9]. In both
of these works the laminated glass was not considered as a perfect sandwich. These
two works, conducted with Saint-Gobain, presented the stretching of the interlayer
following its delamination from glass as the main source of dissipation. Nourry used
a controlled perforation test to evaluate the amount of energy dissipated during the
impact. In this test, the impacting object hits a glass (30 cm width) and is stopped
after a certain distance of penetration. He showed, that almost 87% of the impact
energy is dissipated during impact, after a 14 cm perforation (Figure 1.7).
Decourcelle pursued Nourry’s work and compared the results of the controlled
perforation test with a standard ball drop test through the critical energy needed
to perforate the laminated glass.
Chen et al. presented the interlayer as a polymer bridge between glass shatters
[10] and [11]. In particular, they described the influence of the interlayer on crack
propagation in the glass ply. Namely, the crack propagation velocity was reduced in
the laminated glass compared to a simple monolithic glass under the same impact
conditions.

1.3 Questions
As pointed out in many of the previous studies, for instance by Nourry [8], interlayer
delamination and stretching is the main source of energy dissipation. As explained
in 1.2.2, during impact, because of glass breakage and flexion of the overall structure,
glass shatters go away from each other. In the process, the interlayer is delaminated

9
Figure 1.7: The kinetic energy of the impacting object is plotted as a function of
time and perforation length. This test is conducted at 20 ◦C and the impacting
object has a velocity of 9.2 m s−1 (Results from Nourry’s PhD thesis [8]).

from the glass. In between the delamination fronts, the polymer is stretched (Figure
1.8). Thus we decided to focus on this particular step of the impact.

Figure 1.8: Following the impact, the bending of the plate drives glass shatters away
from each other inducing stretching of the interlayer

Our main objective is to identify the main dissipation processes. We want to

show that the basic understanding presented in 1.2.2 is correct and that the large
amount of dissipated energy is the result of a balance in adhesion. Moreover, we
will quantify the different mechanisms involved in the delamination and stretching
of the interlayer during the impact.
During impact, due to the fast succession of events and to the complicated crack
pattern, it is difficult to target specific dissipating mechanisms. Thus model exper-
iments have been developed. Seshadri and al. have used such model experiments

10
to study the delamination and subsequent stretching of the interlayer. They used a
Through Crack Tensile test (TCT test) to model these two phenomena [12]. This
test presented more thoroughly in 2.5.2, consists in a uniaxial tension test on a
laminated glass in which both glass sides are pre-cracked (Figure 1.9). It partially
reproduces what is happening during the flexion step of the impact in between two
glass parts.

F δ

Figure 1.9: Schematic representation of the Through Crack Tensile Test sample.
The delaminated part of the interlayer is stretched in between the delamination
fronts.

This test combines the actions of the glass/interlayer interface and of the inter-
layer rheology.
Due to the key role of the interlayer, its rheology will be studied. It will provide
a tool to analyze the delamination experiments that will be conducted in a second
time. In this study, we will focus on:

• The rheology of the interlayer in both small and large strains.

• The amount of energy that the interlayer is dissipating.

• The relation between dissipation and polymer chemistry and structure.

Based on this understanding of PVB rheology and with the delamination exper-
iment we will try to answer to the following problems:

• Can we quantify interlayer stretching and delamination velocity?

• How much energy is dissipated in this process?

11
• What is the impact of applied velocity and temperature?

• Can we identify distinct zones of dissipation?

• What is the effect of the level of adhesion?

• Can we provide a basic understanding and modeling of the delamination?

• And finally, what does that imply for the industrial application?

12
Chapter 2

Methods

In this chapter, the preparation of laminated glass samples will be shortly described.
A method to modify the interlayer/glass interface thanks to silanization will be
briefly presented. The different experimental methods (mechanical tests and optical
measurement systems) used to characterize the interlayer behavior and the rupture
of adhesion with the glass will be presented.

2.1 Laminated glass assembly

Laminated glass is generally composed of a layer of Poly(VinylButyral) sandwiched
in between two glass plies. The glass used in this study is a Planiclear glass from
Saint-Gobain with a 2 mm thickness. Laminated glass is assembled at Saint-Gobain
Recherche. At the laboratory scale it is a one day process (Figure 2.1), which
reproduces the industrial process. The glass is first cut using a glass cutter, then
the glass edges are smoothed onto an abrasive band. The PVB interlayer is cut at
the correct dimensions. Assembly is carried out in a clean room after washing the
glass on both sides (industrial concentrated soap highly diluted in ultra pure water),
carefully rinsing and drying the glass. The glass assembly is then put into a vacuum
bag which is subjected to a heat and pressure treatment in an autoclave for 1.5 hour
at 140 ◦C and 10 bar.

2.2 Silanization
For some experiments, the adhesion in laminated glass was modified by silaniza-
tion of the glass surface. In order to obtain a stronger or a weaker adhesion two

13
 Vacuum
bags

PVB film Clean room assembly Autoclaving

140 ◦C/10 bar/1.5 h

Figure 2.1: Schematic representation of laminated glass assembly at the laboratory

scale

(a) APTS (3-Aminopropyl)triethoxysilane (b) OTES triethoxy(octyl)silane

Figure 2.2: The two silanes used for (a) enhancing and (b) reducing the adhesion
between the glass and the PVB interlayer

silanes were used. (3-Aminopropyl) triethoxysilane (APTS) was used to enhance

the adhesion while triethoxy(octyl)silane (OTES) was used to reduce it.

The silanes were applied by a wiping method. This method is not commonly
used in academic research but has proved to be very reproducible. Moreover it is
easily scalable to large samples such as the one used in the peel and TCT tests. A
Cerox
R solution was used to clean the glass. Cerox
R is an abrasive powder which

activates the glass surface by removing molecules already interacting with the glass
silanols (light surface etching of the glass). After rinsing the glass surface with
deionized water the silane solution previously prepared was applied on the cleaned
surface.

The silane agents were pre-hydrolyzed. The silanes were deposited onto the
glass with a cleanroom wiper. Finally, the treated glass was cleaned with a solvent
solution.

More details about the silanization process can be found in the MS degree thesis
of Raphaelle Kulis [13].

14
2.3 Mechanical behavior of the interlayer
The interlayer meachanical behavior was investigated in both the small strain and
large strain regimes. The different devices and techniques used to conduct this study
are presented here. Since the mechanical tests were done on a PVB as received and
that the PVB in laminated glass is subjected to a heat and pressure treatment during
the lamination process, we checked that the behavior in both small strain and large
strain deformation are similar before and after lamination temperature and pressure
treatment. Results can be found in Appendix A.

2.3.1 Dynamical Mechanical Analysis (DMA)

Mechanical response of polymers subjected to small oscillatory strain

When applying an oscillatory deformation on a viscoelastic sample at a given fre-

quency ω ( = 0 cos(ωt)) the response of the material is delayed in time. This delay
δ characterizes the viscoelastic response of the material (σ = σ0 cos(ωt + δ)). One
can then define a complex modulus E ∗ . The real part E 0 , is called the storage mod-
ulus and reflects the elastic storage of the energy. The imaginary part E 00 is called
the lost modulus and accounts for the viscoelastic losses. The ratio between these
two moduli is often denoted tan δ and it is a way to compare the stored versus lost
energy during the deformation of the material. The stress is then given by σ = E ∗ .

Clamp

(t) = cos(ωt) Polymer film sample

Figure 2.3: DMA experiment on a polymer film in uniaxial tension geometry. A

sinusoidal strain is applied on the sample and a shifted sinusoidal stress is measured.
Storage and loss moduli are extracted from there.

In polymers, a relation between time and temperature dependence of E ∗ can

often be found. Thus the high temperature behavior corresponds to the behavior
at low frequency or at long time. Similarly, the behavior at low temperature corre-
sponds to the behavior at high frequency or at short time. This relation between
time and temperature is often used to build master curves of the mechanical be-

15
havior of the polymer (storage and loss moduli) from experiments made at different
frequencies or temperatures. This enables, in the end, to get data on a much larger
range of temperatures and frequencies. The shift factors on the frequency axis (usu-
ally called aT ) for each curves follow specific laws that enable to predict the behavior
at higher or lower temperature or frequency. These coefficients are defined as the
ratio between the frequency at temperature T and the frequency at the reference
temperature Tref (equation 2.3.1).

ωT
aT = (2.3.1)
ωTref
Thanks to these shifting coefficients, the storage modulus E(ω, T ) at frequency ω
and temperature T can be shifted at the corresponding frequency ω 0 for the reference
temperature Tref with the relation of equation 2.3.2.

E(ω, T ) = E(ωTref , Tref ) = E(ω/aT , Tref ) (2.3.2)

The Arrhenius law is a first example of shifting coefficient law (equation 2.3.3).
In this equation R is the universal gas constant and Ea is an activation energy. This
type of law can be used for example to describe hydrogen bond dynamics.

Ea 1 T
ln(aT ) = ( − ) (2.3.3)
R T Tref
William, Landel and Ferry [14] have shown that in the case of polymers, the
time dependence and the temperature dependence are related through the so called
WLF law which gives shifting coefficient aT linking time to temperature at a given
reference temperature as:

−C1 (T − Tref )
log(aT ) = (2.3.4)
C2 + T − Tref
This equation is a semi-empirical relation that Wiliam et al. have found from
experiments on different polymers. C1 and C2 are often said to be universal coeffi-
cients. Typical values of these coefficients are C1 = 17.4 and C2 = 51.6◦C. However
these values do not apply to many other polymers.

Experimental setup

Dynamic mechanical analysis was conducted on a TA Instruments Q800 apparatus.

As PVB is a soft thin film, the tension geometry was used (Figure 2.3). Small pieces
of precise dimension around 15 mm length and 10 mm width were used. A preload

16
of 0.01 N is applied to have a straight sample between clamps and prevent buckling
due to clamping.
The linear regime extends at least up to 0.5%. In order to always be in the linear
regime of strain and to remain at low stresses, the strain in DMA measurements
will be kept at a value of 0.01% which will also be used in the rheometer (in which
higher strains will induce stresses out of the scope of the rheometer, even at room
temperature).

2.3.2 Rheology
DMA measurements can only cover a temperature range over which the material is
solid and does not flow. In our case the glass transition temperature being around
25 ◦C to 35 ◦C, measurements above 60 ◦C could not be made and results between
40 ◦C and 60 ◦C have to be regarded with great care as they might be subjected to
large deviation. In order to compare polymers in a temperature range above the
glass transition a rheometer with a parallel plates geometry was used.
The apparatus is an Anton-Paar rheometer (Physica MCR 501). The parallel
plates geometry used was made of a small bucket of 5 cm diameter to contain the
sample and a disk of the same diameter to apply a controlled oscillatory deformation.
In order to get a good measurement three layers of polymer were put inside the
small bucket which was used as a mold and kept at 160 ◦C during 30 min before
measurement. The melted layers of polymer then form a single volume of material
thick enough to have a measurable value of stress.
Contact between the upper plate of the geometry of the rheometer and the sample
was kept by imposing the normal forces to be zero at all time during the experiment
(so that contact is not lost if the sample shrinks). The temperature range was 40 ◦C
to 160 ◦C with 5 ◦C steps. At each temperature step, a frequency range from 0.1 Hz
to 10 Hz was covered. Before doing any measurement the temperature was kept
constant (with a precision of 0.2 ◦C) during at least 10 min.

2.3.3 Uniaxial tension

Uniaxial tension experiments were conducted on a Zwick (Hamsler HC25) hydraulic
machine. A force cell of 1 kN was used. Displacements were measured thanks to both
the clamp displacement and a video camera Baumer BM20. In order to control the
temperature a closed cabinet was used around the clamps. Cooling was obtained
by liquid nitrogen and heating by an electric resistance. Thanks to a PID ther-

17
moregulator 2216L from Eurother Automation, the temperature was kept constant
close to a user defined temperature. The homogeneity of the temperature inside the
cabinet was checked several times in different places of the cabinet. Humidity was
not controlled and it was assumed to be identical to the ambient humidity (in a
range from 20 to 50 HR).
Low force measurements were conducted on an Instron machine (Instron 5565)
with a load cell of 100 N. Temperature was controlled in a closed cabinet using a
2216L Eurother Automation thermoregulator. Displacement was monitored with a
video extensometer: two white dots are placed on the sample and followed automat-
ically by the extensometer.
The tests were conducted at controlled displacement rate on shouldered test bars
of controlled dimensions (Figure 2.4).

10 mm

4 mm
20 mm

5 mm

Figure 2.4: Uniaxial tension sample shape

2.4 Different models to describe the constitutive

behavior of the interlayer
The complex behavior of the interlayer described later in Chapter 3, is difficult to
model as it displays simultaneously a viscoelastic behavior, a hyperelastic behavior
and a plastic behavior. We choose to use a Generalized Maxwell description for the
viscoelastic part [15] through a Prony’s serie. It will model the time and temperature
dependence of the material. As the polymer is subjected to large strain, linear
elasticity is no longer adapted. The hyperelastic model will also have to take into
account of the hardening of the material at really high strains. Thus an Arruda-
Boyce model will be used [16]. These different models are described here.

18
2.4.1 Small strain description: Generalized Maxwell model
In order to model the small strain behavior one can use a generalized Maxwell
model and the associated Prony’s serie. In this model (Figure 2.5), the material
is described by dashpots to represent elastic relaxation with different characteristic
times (τi ) of the material and the associated elasticity with springs (Ei ). The springs
and dashpots are associated in series for each characteristic time. All the branches
are in parallel. A final spring is associated in parallel with these dissipative branches
to represent the long term elasticity (E∞ ).

... ...
En Ei E1
E∞
τn τi τ1
... ...

Figure 2.5: The generalized Maxwell rheological model

The global time dependent moduli describing the generalized Maxwell rheological
model can be written either in a time or frequency base. In the time base, the
relaxation function of the Young modulus can be written as a Prony’s serie:

N
E(t) = E∞ + Ei . exp(−t/τi ) (2.4.1)
X

i=1

In the frequency base the Young modulus can be decomposed in the storage and
loss components. In order to do so we will define the instantaneous modulus E0
which is given by E0 = E∞ + Ei which corresponds to immediate response of
PN
i=1
the material. The coefficient ai are also defined as ai = Ei /E0 . Finally the frequency
base storage and loss moduli can be defined by the following expressions:

N
ai τi2 ω 2
E (ω) = E∞ + E0
0
X

i=1 1 + τi ω
2 2

N
ai τ i ω
E 00 (ω) = E∞ + E0 (2.4.2)
X

i=1 1 + τi ω
2 2

2.4.2 Hyperelasticity: Arruda-Boyce model

The Arruda-Boyce model is used to model large strain hyperelasticity including the
strain hardening displayed at very large strains. This model is based on a statistical

19
description of the polymer network. In a uniaxial tension test on an incompressible
material, the stress expression is the following:

1 X
 5
iCi I1i−1

σ = 2µ λ − 2 (2.4.3)
λ i=1 λ2(i−1)
m

where µ is the shear modulus, λ is the stretch applied on the material and I1 is
the first invariant I1 = λ21 + λ22 + λ23 . The Ci are the coefficients of the Langevin
function: C1 = 1/2, C2 = 1/20, C3 = 11/1050, C4 = 19/7000 and C5 = 519/673750.
λm is the maximal chain extensibility. A compilation of several results found in the
literature are put together in Appendix B to explain this expression.

2.5 Delamination experiments on laminated glass

2.5.1 Peel test
In order to assess the adhesive properties of the interlayer on the glass, some peel
experiments were conducted by Raphaelle Kulis [13] during her internship. Special
peel samples were prepared. A glass of 5 cm width, 15 cm length and 2 mm thickness
were used. The interlayer assembled with the glass was 2 cm width, 20 cm length
and 0.76 mm thickness.
The assembly (Figure 2.6) was made as for a classical laminated glass (cleaning
and assembly in clean room) but instead of applying a second glass on top of the
interlayer a cloth backing was put above the interlayer. A small part of the bottom
glass was also covered with Kapton
R to make the peeling initiation easier.

1 cm

5 cm 2 cm

15 cm
(a) Top view

Figure 2.6: Laminated assembly for peel test [13].

The peel sample was then cured in the autoclave using the standard procedure
for laminated glass.
The peel sample was tested on a Zwick Z010 testing machine with a 1 kN load
cell. The peeling occurs at 90◦ . The laminated peel sample is put on rolls which

20
transform the vertical motion of the upper clamp into a horizontal sliding of the
glass (Figure 2.7). The peel angle is constant through the test. The displacement
speed is controlled.

Glass

Upper clamp (rolls)

PVB + Stiff backing

Lower clamp

Figure 2.7: Schematic view of the peel setup. Two rolls enable the peel angle to be
kept constant through the test equal to 90◦ [13].

2.5.2 Through crack tensile test

The through crack tensile test is a uniaxial tension test made on a pre-cracked
laminated glass (Figure 2.8). The test is conducted on a Zwick Hamsler HC25
hydraulic machine with a 10 kN load cell. The laminated glass sample used for this
experiment has a 5 cm width and a 10 cm length. The two glass panes are 2 mm
thick and the interlayer thickness is most of the time 0.76 mm. Just before the TCT
experiment, the two glasses are cut in their middle through their width. A first
scratch is made on the glass along its width thanks to a diamond cutting wheel.
Then the laminated sample is subjected to very small bending in order to propagate
the crack. The sample is then mounted in the tensile rig and a pressure of 8 MPa
is applied in-between the clamps to avoid any sliding of the sample. Finally the
cracked laminated glass is subjected to a uniaxial tension experiment. During the
experiment, the upper clamp velocity is controlled and the bottom glass is fixed.
The interlayer is delaminating from the glass and stretched. The position of the
delamination fronts is measured with a Baumer BM20 camera. The environment is
controlled as in 2.3.3.

21
 δ

0.76 mm
2 mm

Glass

10 cm Crack

PVB

5 cm

Figure 2.8: Schematic representation of the Through Crack Tensile Test sample.
During the test, the displacement δ and the associated velocity δ̇ are controlled.

2.6 Optical methods

2.6.1 Video acquisition

Video acquisition for displacement measurements were conducted with a Baumer

BM20 industrial camera. As there was no software to enable movie acquisition a
small acquisition software was developed with Matlab. Matlab Video Acquisition
toolbox was used to acquire the images and to control the camera. A gui was setup
to enable a better usability by other users. The software has been packaged with
Matlab App Tools and is available at this address:

http://www.mathworks.com/matlabcentral/fileexchange/46879-gigeacq-m

This code is free to use and to distribute under BSD licensing. The code uses
two functions provided by Mathwork Central named CalculateFrameRate and Cal-
culatePacketDelay which can be found here:

http://www.mathworks.com/matlabcentral/answers/uploaded_files/6148/CalculateFrameRate.m

and there

http://www.mathworks.com/matlabcentral/answers/uploaded_files/1061/CalculatePacketDelay.m

A Photon Mini UX100 fastcamera was also used for digital image correlation
measurements and for sensitivity purpose in the photoelastic measurements.

22
2.6.2 Digital image correlation
Digital image correlation was sometimes applied to the TCT test. Ink was sprayed
on one side of the PVB interlayer. A second interlayer film was added on top of the
first one (Figure 2.9). Thus, the ink was not in contact with the glass to prevent a
change in adhesion. Moreover, as the dots are on the middle plane they did not get
out of the focal plane during the experiment.

Figure 2.9: Schematic view of the digital image correlation sample (left) and exper-
iment (right). Paint dots are randomly spread to obtain a speckle, in the middle
plane of the PVB layer.

The image acquisition focused on a small area on the bottom glass (which is
fixed during the experiment 2.5.2). A long range microscope from Questar (Questar
Q100) was used.
During the experiment (Figure 2.10) displacements were measured along a verti-
cal line (white line). These displacements are corrected by the average displacement
of some points which are supposed to be fixed (white square). This will subtract
the rigid body displacement due to unwanted movements of the machine during the
test. At some point of the experiment, these reference points start to move and
they can no longer be used as reference to correct the machine displacements. Thus
a mean value of their displacements during the last 10 frames are used to evaluate
future corrections. At the end of the experiment these corrections are less important
as the machine displacements are less pronounced.

2.6.3 Photoelasticity
Photoelastic measurements were conducted on both laminated glass samples during
the TCT test and the uniaxial tension of the interlayer alone. Two cross polarizers
and two quarter wave plates were placed around the sample as described on Figure

23
 Vertical line:
Measure of displacements
Delaminated interlayer

Delamination front
on front glass

Delamination front
on back glass

Area used for the

correction of displacement
500 µm
Adhered interlayer

Figure 2.10: Digital image correlation frame. The displacements measured on the
vertical white line are corrected by the displacements (supposed to be due to exper-
imental noise) of the white region out of the delaminated area.

2.11. The first polarizer is simply called the polarizer while the second one is referred
to as the analyzer.
Some materials exhibit an anisotropy of the refractive indices. These materials
are called birefringent. As the optical properties of the material are related to their
molecular organization (in particular in polymers to chain alignment), birefringence
appears when a stretch is applied to a transparent material.
A simple way to visualize this birefringence is to illuminate with a white light
and to put a polarizer and an analyzer on each part of the sample. If the principal
optical directions of the polarizer and of the analyzer are crossed at 90◦ from each

Crossed
Polarizer Analyzer
Light
Camera

/4 /4

Figure 2.11: Photoelasticity measurement: two cross polarizers and two quarter
wave plates were placed around a PVB uniaxial tension sample or laminated glass
TCT test sample.

24
other, one can observe two things (Figure 2.12):

• Some completely dark bands called isoclinic bands

• Some colored bands called isochromatic bands

Colored isochromatic
fringe

Dark isoclinic fringe

Figure 2.12: A photoelastic pattern observed on a PMMA sample in-between cross

polarizers in white light. One isoclinic dark line goes through the sample indicating
the principal stress direction. The isochromatic colored bands are regions of identical
chain alignment.

The isoclinic bands correspond to the part of the sample where the stretch is
aligned with the polarizer and thus polarized light is completely cut off by the
analyzer (no matter the wavelength). These bands will be parallel to the principal
strain or stress directions. If the crossed polarizer/analyzer are rotated, the position
of these line will change. The isochromatic bands corresponds to regions over which
the light phase was shifted from a multiplicity of the light time period. For each
wavelength, the refraction will be different and one observes a color pattern. Each
color bands corresponds to an area where there is an equal amplitude of chain
alignment or local strain. Unlike isoclinic bands, the isochromatic bands do not
depend on the polarizer/analyzer orientation. Quarter wave plates are used in order
to get a circular light polarization. This will remove the isoclinic fringes which do
not bring much as the principal stress directions are well known in the uniaxial
tension tests that we are conducting. Moreover, monochromatic light is used to
only get variation of the light intensity over the sample. The areas with the same
intensity present the same amplitude of chain alignment.
In the case of a monochromatic light (wavelength λ0 , the Gaussian network
theory description of photoelasticity described by Treloar ([17]) gives the intensity
variation as a function of the stretch applied to the sample:

25
 2Cπ(n1 − n2 )t 2C 0 π(σ1 − σ2 )t
" # " #
I ∝ cos 2
≈ cos2 (2.6.1)
λ0 λ0
where n1 and n2 are the principal optical coefficient, and σ1 and σ2 are the
corresponding principal stresses. C and C 0 are constants related to the network
properties. t is the interlayer thickness which also changes during the test and that
has to be accounted for. In the case of an incompressible material subjected to
uniaxial tension this expression leads to:

2C 0 π(λ2 − λ1 )t
" #
I ∝ cos 2
(2.6.2)
λ0
Finally, we can get from this equation that a complete light extinction will be
observed periodically. The order of the extinction or birefringent order is given by:

C(λ2 − λ1 )
N= t (2.6.3)
λ0

2.7 Differential Scanning Calorimetry

Differential Scanning Calorimetry experiments were conducted on Q200 apparatus
from TA-Instruments in an hermetically closed aluminum pane. Measurements are
done with a nitrogen flow of 50 mm min−1 .

2.8 X-ray scattering

Small angle X-ray scattering experiments were conducted for us by Quentin De-
massieux. Shouldered test bar samples with an effective length of 20 mm were used.
The tests were conducted at the Advanced Photon Source (Argonne National Lab-
oratory) 5ID-D, B run beam. During the scattering exeperiments, the sample was
loaded with a Linkam TST350 device which is composed of a traction machine en-
closed in a oven equipped with a heating element in contact with the sample to
control the temperature. The heating element is transparent to X-rays. The dis-
placement was only measured thanks to clamp displacement as no images could be
taken during the test. Samples were subjected to a uniaxial tension at a strain rate
of 0.01 s−1 . X-ray scattering was recorded during this time. Only SAXS results will
be presented here.
The 5ID-D Run B beam has the following properties:

26
Energy (keV) Wavelength (nm) SAXS distance (m) SAXS range Q (nm−1 )
9.8 0.127 8.505 0.013–0.78

Table 2.1: Parameters of the beam and detector for the 5ID-D Run B beam.

27
28
Chapter 3

A complex structure and rheology

3.1 Introduction
The relationship between the mechanical behavior of the interlayer and its chemical
structure is of prime interest in order to understand the dissipation mechanisms
occurring during impact. Indeed, the delaminated part of the interlayer will be sub-
jected to large stretch at various strain rates. We will first present here the chemistry
of the PVB interlayer. Then the small strain mechanical response of the interlayer is
investigated to characterize its time and temperature dependence. The large strain
uniaxial cyclic tests provide a better understanding of the dissipation mechanisms
at different strain rates and temperatures. Since during the delamination, the inter-
layer is subjected to a large level of strain, it is necessary to investigate both regimes
to build a proper description of the interlayer behavior. Finally, photoelastic mea-
surements will help us to link these dissipation processes to the chemical structure
of the interlayer through a simple rheological model.

3.2 Poly(Vinyl Butyral) interlayer

3.2.1 Chemistry
PVB is a random polymer chain made of three monomers (Figure 3.1). The synthesis
process goes through three main steps [18]. The first one is the polymerization of
vinylacetate in a chain of poly(vinylacetate). This chain is then hydrolyzed which
leads to the formation of poly(vinylalcohol). Poly(vinylalcohol) finally reacts with
butyraldehyde to form butyral cycles. The final composition of the chain which
presents the optimal adhesion and mechanical properties is close to 1 or 2 wt% of

29
acetate groups, 18 wt% of alcohol groups and 80 wt% of butyral cycles.

    




Vinylacetate Vinylalcohol Vinylbutyral

Figure 3.1: The three monomers found in the PVB structure.

The polymer weight found from Size Extrusion Chromatography is of 197 kg mol−1
(relative to polystyrene calibration) with a polydispersity index of 1.42.
The industrial product contains also a plasticizing agent which is expected to
decrease the glass transition temperature of the polymer from 70 ◦C to close to ambi-
ent temperature. The plasticizer is generally triethylene glycol di-(2-ethylhexanoate)
(Figure 3.2). The amount of plasticizer is up to 20wt% leading to a glass transition
temperature around 25 ◦C to 30 ◦C.


 
 


Triethylene glycol bis(2-ethylhexanoate)

Figure 3.2: Plasticizer molecule.

3.2.2 Hydroxyl groups

This polymer contains a certain amount of hydroxyl (OH) groups that are available
for bonding to the glass but also for forming inter-chain bonds. Polarized hydrogen
atoms are linked to a more electronegative atom such as oxygen. In the case of PVB
it can be an hydrogen on the vinylalcohol group, the silanol group at the glass surface
or any other OH group. Hydrogen bonds form between this polarized hydrogen and
an electronegative atom which can be here the acetate group. This interaction has a

30
strength around 10 kJ mol−1 which has to be compared to the covalent bond which
has an energy > 100 kJ mol−1 and the Van der Waals interaction which has an energy
around 1 kJ mol−1 . However, a large number of this relatively weak hydrogen bonds
can form strong links. The number of hydroxyl groups and acetate groups are thus
very important for the rheology and mechanical behavior of the polymer but also
for the adhesion strength. The PVB reference which has been used during this work
presents an amount of OH groups around 17-18wt% or 40mol%.
By changing the hydroxyl group content, both adhesion on glass and bulk ma-
terials properties are impacted. In previous works made in Saint-Gobain ([19] and
[18]), a relationship between interlayer chemistry (and chemical structure) and its
mechanical behavior has been studied. In particular it has been shown that the
amount of hydroxyl bond has a strong influence on the mechanical behavior of the
interlayer. For example Klock ([19]) has shown that the Young modulus of the
PVB is strongly affected by the amount of hydroxyl groups (Figure 3.3). In this
figure, the amount of hydroxyl groups varies between 40 mol% and 85 mol% and
the young modulus increases from 1 MPa to 1 GPa. The plasticizer amount is fixed
at 26 mol%. The polymer is going through the glass transition towards the glassy
plateau as the amount of hydroxyl groups increases at constant temperature and
constant plasticizer content. The hydroxyl group content is also changing adhesion
properties. In Figure 3.4, the peeling energy is measured during a 90◦ peel test at
10 mm min−1 , at room temperature and shows this large increase.

Figure 3.3: Young modulus as a function of the amount of OH groups. The OH

groups content impacts the rheological properties of the material [19].

Mertz [18] has also shown that the hydroxyl groups form hydrogen bonds which
break as the interlayer is stretched. Indeed he observed by FTIR that the ratio of
free hydroxyl groups over hydroxyl groups forming hydrogen bonds, is decreasing as

31
Figure 3.4: The peel energy as a function of the OH group amount. An increased
value of the peeling energy is found for higher hydroxyl group content [19]. This
increase is highly non linear.

the interlayer is stretched.

Finally, Mertz and Klock have shown a similar impact on the mechanical behavior
of the amount of plasticizer. This plasticizer is supposed to prevent the formation
of the hydrogen bonds in between the polymer chain by spacing the chains. Thus
as the plasticizer amount increases, the interlayer becomes softer. For a plasticizer
amount above 20 mol%, a PVB with a hydroxyl content of 43 mol% is in a rubbery
state.

3.3 Rheology of the PVB

In order to study the mechanical behavior of the PVB interlayer, in both linear (small
strain) and non-linear (large strain) regimes, rheology experiments (namely DMA
and parallel plate rheometer) and uniaxial tension tests were conducted. Hooper
made similar mechanical tests on the same PVB interlayer [20]. Others, used differ-
ent PVB but provided complementary insight such as dynamic loading experiments
at large strain [21] or large strain cyclic experiments [22]. In the next sections we
present our own results which confirm the literature and also provide a new insight
in the relationship between the mechanical behavior, the chemical structure and the
energy dissipation in the PVB.

32
3.3.1 Small strain viscoelasticity
In order to investigate the small strain behavior of the interlayer and to get a proper
description of the time dependence of its linear behavior, DMA and parallel plate
rheology experiments were conducted. The DMA experiments measured the time
and temperature dependence of the young modulus in a range of temperature from
−40 ◦C to 60 ◦C. The rheometer experiments give the dependence on time and
temperature of the shear modulus in the range of temperature from 40 ◦C to 160 ◦C.
We also checked that the autoclaving process and the delamination from glass
after the autoclave did not affect the interlayer behavior. Thus, the mechanical
behavior of the interlayer can be measured on the PVB film as received (Appendix
A).

Dynamical mechanical analysis

DMA experiments were conducted first at a frequency of 1 Hz with a strain am-

plitude of 0.01%. Temperature increases from −40 ◦C to 80 ◦C at a heating speed
of 3 ◦C min−1 . The glass transition temperature of the plasticized PVB is around
30 ◦C (Figure 3.5). To compare with other experiment it is important to note that
the measured glass transition varies with the heating speed. The polymer displays
a glassy behavior for temperatures lower than 20 ◦C. No clear value for the glassy
storage modulus can be measured there, as it keeps increasing from 0.3 GPa to 1 GPa
as the temperature decreases from 10 ◦C to −40 ◦C. On this plateau, the dissipation
ratio is low as the loss modulus represents only 10% of the storage modulus. At
temperatures higher than 50 ◦C, the polymer exhibits a rubbery behavior. The stor-
age modulus drops to values ranging from 0.8 MPa to 1 MPa. Again, the dissipation
ratio is low in this regime, where the loss modulus represents only 10% to 20% of
the storage modulus. However, during the glass transition, the dissipation ratio is
maximal (tan(δ) is higher than 1). There seems, as well, to be a slight increase of
the dissipation at 70 ◦C but as the DMA reaches its limits, the result has to be taken
with caution.
As we are interested in the time dependence of the PVB behavior, the same
DMA experiment is made at different temperatures from −40 ◦C to 60 ◦C but for
different oscillation frequencies. Each 5 ◦C, 5 frequencies per decades from 0.1 Hz
to 10 Hz are tested (Figure 3.6(a)). Thanks to the time/temperature equivalence
principle, a master curve can be obtained by shifting the different curves. A reference
temperature of 20 ◦C was used to build the curve.
In Figure 3.6(c), the storage, the loss and the dissipation ratio master curves are

33
 103 1.2
Storage Modulus
Loss Modulus
tan(δ) 1
102
Young Moduli (MPa)

0.8

tan(δ)
101
0.6

0.4
100

0.2
10−1
0
-40 -20 0 20 40 60 80
Temperature (◦ C)

Figure 3.5: DMA experiment on PVB at 1 Hz and 0.01% deformation for tempera-
ture ranging from −40 ◦C to 80 ◦C.

plotted together. These results are similar to the one found by Hooper on the same
interlayer [20].
The logarithmic values of the shift coefficients at this temperature can be plotted
as a function of the inverse of the temperature (Figure 3.7). This can be used to
extract WLF coefficients which are here C1 = 81 and C2 = 417 ◦C.
If we want to expand our description of the time dependence of the PVB behavior
at higher temperature, DMA is no longer suited as the polymer film becomes too
soft and deforms under its own weight in the traction DMA setup. That is why a
parallel plate rheometer was used to screen higher temperatures or longer time.

Parallel plate rheology at higher temperatures

Experiments were performed in the parallel plate geometry on three superimposed

PVB films. At each 5 ◦C step for temperatures ranging from 35 ◦C to 160 ◦C, fre-
quencies from 0.1 Hz to 10 Hz were screened. A master curve of the shear moduli
can be built from these results (Figure 3.8) with a reference temperature of 40 ◦C
and the coefficients are displayed in Figure 3.9. The time/temperature coefficients
still follow a WLF law. However large vertical shift coefficients have to be applied
away from the glass transition. The WLF coefficients for this master curve are C1 =
12 and C2 = 74◦C.

34
 Storage Young Modulus (MPa)
Storage Young Modulus (MPa)

-40◦ C -40◦ C
-30◦ C -30◦ C
-20◦ C -20◦ C
102 -10◦ C -10◦ C
102
0◦ C 0◦ C
10◦ C 10◦ C
20◦ C 20◦ C
101 30◦ C 30◦ C
40◦ C 101 40◦ C
50◦ C 50◦ C
60◦ C 60◦ C
100
100
10 −1
100
10 1
10−3 100 103 106 109 1012
Frequency (Hz) Frequency (Hz)
(a) Raw data (b) Time/temperature superposition

103
1

102
Young Moduli (MPa)

0.8

Storage Modulus 0.6

tan(δ)
101 Loss Modulus
tan(δ)
0.4

100
0.2

10−1 0
10−3 100 103 106 109
Frequency (Hz)
(c) Storage and loss young moduli with the dissipation ratio tanδ at 20 ◦C

Figure 3.6: DMA experiment on PVB for frequencies ranging from 0.01 Hz to 1 Hz,
0.01% deformation and for temperature ranging from −40 ◦C to 60 ◦C. In (a) the raw
data concerning the storage modulus. In (b) the Time/Temperature superposition
for the storage modulus. The reference temperature was set at 20 ◦C. In (c), the
loss modulus and the dissipation ratio are added. Maximal dissipation occurs at the
glass transition between 0.1 Hz and 1 Hz at this temperature.

35
 0.6
Experimental results
0.4 WLF fit

0.2
log(aT )

-0.2

-0.4

-0.6
-0.1 -0.05 0 0.05 0.1
1/(T-Tref ) C ◦ −1

Figure 3.7: DMA shift coefficient at a reference temperature of 40 ◦C (blue) and

WLF law fit with (C1 = 19 and C2 = 143 ◦C) (red).

Storage Modulus
100 1
Loss Modulus
tan(δ)
Shear Moduli (MPa)

0.8
tan(δ)

0.6
10 −1

0.4

0.2

10−2 0
10−8 10−7 10−6 10−5 10−4 10−3 10−2 10−1 100 101
Frequency (Hz)

Figure 3.8: Rheometer shear modulus measurements in plane/plane geometry with

0.01% deformation, temperature ranges from 40 ◦C to 160 ◦C and frequency from
0.1 Hz to 1 Hz. Reference temperature is 40 ◦C.

36
 1.5

0.5

log(aT )
0

-0.5

-1

-1.5
-0.2 -0.1 0 0.1 0.2
1/(T-Tref ) C ◦ −1

Figure 3.9: Rheometer experiment, shift coefficients following a WLF law for the
master curve at 40 ◦C (C1 = 12 and C2 = 74 ◦C).

From a time/temperature master curve, the mass between entanglements can

be deduced from the value of the storage modulus at the frequency corresponding
to the minimum of the loss modulus curve, on the rubbery plateau. The relation
between the mass between entanglements Me and the storage modulus G0 is given
by:

ρRT
G0 = (3.3.1)
Me
The material displays a rather large rubbery plateau from which we can ex-
tract an order of magnitude of the polymer chain length between entanglements:
Me = 7 kg mol−1 which can be compared to the approximate total chain length
200 kg mol−1 . This two values give 30 entanglements per chain. This is quite a high
amount of entanglements which might also contribute to the difficulties encountered
in the time/temperature superposition by the WLF law.

Complete description from −40 ◦C to 160 ◦C in small strain

The DMA and rheometer results can be combined in a single curve (Figure 3.10).
As the rheometer gives shear modulus and the DMA gives a Young modulus, to
combine the two curves, we had to assume that the material is incompressible so
that G = E/3. As most of our experiments will be done at 20 ◦C we made this
master curve at a reference temperature of 20 ◦C even if it is a little bit below the
glass transition temperature. Note that this temperature is also out of the range

37
of the rheometer experiment. For the rheometer experiment alone the master curve
was previously built at 40 ◦C and is now shifted at 20 ◦C for the superposition with
DMA experiment.

103
1

102
Young Moduli (MPa)

0.8

Storage Modulus 0.6

tan(δ)
101
Loss Modulus
tan(δ)
0.4
100

0.2

10−1
0
10−8 10−6 10−4 10−2 100 102 104 106 108 1010
Frequency (Hz)

Figure 3.10: DMA and rheometer can be grouped to form a complete small strain
description from the pHz to the THz. Reference temperature is 20 ◦C.

With a generalized Maxwell model, a numerical description of the previous mas-

ter curves can be used to fit the experimental results (Figure 3.11). A 23 terms
Prony series is used. The Prony series coefficients are also given in Table 3.1.

3.3.2 Large strain uniaxial tension

The small strain description of the interlayer behavior is not sufficient to describe
the behavior during impact as the interlayer is deformed at strain higher than 100%.
Large strain uniaxial tension experiments were conducted.
Very few results were found in the literature giving cyclic experiment on the
PVB. Only Seshadri [22] made two cyclic experiments on a PVB interlayer (not
exactly the same as ours) at two different strain rates. His results are similar to the
one we have here. The cyclic uniaxial tension is of utmost importance in this work
as it gives an estimation of the dissipated energy when the polymer is stretched to
a certain level of deformation. The amount of dissipated energy by the polymer is
directly related to the impact performance of the laminated glass as emphasized by

38
 Storage Modulus (MPa)

102
Experiment
Model

100

10−10 10−5 100 105 1010

Frequency (Hz)
(a) Storage modulus
Loss Modulus (MPa)

101
Experiment
Model

100

10−1

10−10 10−5 100 105 1010

Frequency (Hz)
(b) Loss modulus

Figure 3.11: The small strain behavior can be completely described by a 23 terms
Prony serie. It will describe both the storage (a) and the loss (b) moduli on these
master curves at 20 ◦C.

39
 ai τi ai τi
1.52e-01 1.00e-11 6.07e-02 1.00e+00
7.59e-02 1.00e-10 3.79e-02 1.00e+01
7.59e-02 1.00e-09 7.59e-03 1.00e+02
7.59e-02 1.00e-08 1.52e-03 1.00e+03
7.59e-02 1.00e-07 2.28e-04 1.00e+04
7.51e-02 1.00e-06 2.28e-04 1.00e+05
7.48e-02 1.00e-05 2.66e-04 1.00e+06
7.44e-02 1.00e-04 2.66e-04 1.00e+07
7.44e-02 1.00e-03 3.04e-04 1.00e+08
6.83e-02 1.00e-02 2.28e-04 1.00e+09
6.83e-02 1.00e-01 7.59e-05 1.00e+10
1.52e-05 1.00e+11

Table 3.1: The Prony series coefficients for the global DMA/Rheometer master curve
at a reference temperature of 20 ◦C.

Nourry [8].

Effect of time and temperature

Uniaxial tension tests were performed on PVB shouldered samples at different strain
rates and five temperatures 10 ◦C, 20 ◦C, 30 ◦C, 50 ◦C and 70 ◦C. This set of experi-
ments provides a picture of the large strain behavior from the glassy to the rubbery
regime through the glass transition. The different curves can be found in Figure
3.12. In these tests the polymer is stretched to 200% and unloaded at the same
strain rate. Different behaviors are observed depending on temperature and applied
strain rate:

• At low temperature, 10 ◦C (3.12(a)) or at the higher strain rates (1 s−1 ), at

20 ◦C (3.12(b), the interlayer presents a plastic like behavior. When the strain
reaches approximately 10%, the stress is constant up to a strain of 40% as is
observed for plastic materials. Moreover, the initial modulus presents a slight
increase with the strain rate from 70 MPa to 200 MPa when the strain rate
is going from 0.01 s−1 to 1 s−1 as the material is on the glassy plateau. This
behavior is close to the one observed for highly crosslinked polymers.

• At 30 ◦C, the behavior is close to the one observed for loosely crosslinked
polymers. It presents a viscoelastic behavior with a strain hardeing when the
strain is higher than 100%. The material is in a rubbery state and presents
a soft initial modulus which ranges from 2 MPa to 8 MPa as the strain rate
increases from 0.01 s−1 to 1 s−1 .

40
 • At high temperature 50 ◦C (3.12(d) and 70 ◦C (3.12(e)), the behavior is still
similar to the one observed for loosely crosslinked polymers. However, it is
important to notice that at 50 ◦C the interlayer presents a large permanent
deformation (about 20%) which increases at 70 ◦C (about 50%). This is the
result of polymer flow induced at these temperatures by the stretching of the
sample. The stress level decreases also a lot: by a factor 4 between 30 ◦C and
50 ◦C and a factor 10 between 50 ◦C and 70 ◦C. This important change in the
stiffness of the material was not observed in the small strain experiments.

• At 20 ◦C the range of strain rates is large enough to go through the glass

transition. Indeed the higher strain rate is similar to the plastic behavior
observed at 10 ◦C and the lower strain rate is close to the rubbery behavior
observed at 30 ◦C.

The different curves present a stiff hardening of the stress when the strain in-
creases above a threshold value (m ). The threshold value, to this stiff hardening
regime, increases with increasing temperatures and decreasing strain rates. At 10 ◦C
the hardening begins around 40% strain. At 20 ◦C the hardening begins at 50%
strain at 1 s−1 and only at 100% for the lowest strain rate 0.001 s−1 . For the higher
temperatures (above the glass transition), the hardening starts in between 100%
and 150%. For the highest temperatures and slowest strain rates, the hardening is
no more visible on the curves.

At high strain rates and low temperatures, the glassy modulus is about 200 MPa
and on the opposite side, at low strain rates and high temperatures, the rubbery
modulus is around 0.3 MPa. These moduli are comparable to those found in DMA
experiments. Moreover, we can build a master curve of the initial modulus found in
large strain uniaxial tension by shifting the results found at different temperatures
on the reference temperature 20 ◦C (Figure 3.13). This plot does not display the
modulus as a function of the frequency as in the DMA master curves. However,
it shows that the glass transition is occurring here at a temperature lower than in
DMA. Indeed, at 20 ◦C, the initial modulus has already decreased more significantly
than in the DMA experiment. This result might be explained by the fact that the
polymer chain movement is eased by the large strain applied on the material. The
glass transition might be shifted to lower temperature by the application of large
strains.

41
 70 80
˙ = 1s−1 ˙ = 1 s−1
60 ˙ = 0.1s−1 ˙ = 0.1 s−1
True Stress σT (MPa)

True Stress σT (MPa)

˙ = 0.01s−1 ˙ = 0.01 s−1
50 60
˙ = 0.001 s−1
Yield stress
40 Plastic behavior Strain hardening start
40
30
20
20
10
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Nominal Strain N Nominal Strain N
(a) 10 ◦C (b) 20 ◦C

25 5
˙ = 1s−1 ˙ = 1s−1
˙ = 0.1s−1 ˙ = 0.1s−1
True Stress σT (MPa)
True Stress σT (MPa)

20 ˙ = 0.01s−1 4 ˙ = 0.01s−1

15 3

10 2

5
1

0
0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Nominal Strain N Nominal Strain N
(c) 30 ◦C (d) 50 ◦C

˙ = 0.1s−1
0.6 ˙ = 0.01s−1 Viscous Flow
True Stress σT (MPa)

˙ = 0.001s−1
0.5

0.4
Permanent deformation
0.3

0.2

0.1

0
0 0.5 1 1.5 2
Nominal Strain N
(e) 70 ◦C

Figure 3.12: Uniaxial tension test at controlled displacement speed and different
temperature: 10 ◦C (a), 20 ◦C (b), 30 ◦C (c), 50 ◦C (d) and 70 ◦C (e). Up to 150% or
200% of maximal deformation. Loading and unloading are at the same strain rate.

42
 103
70◦ C
50◦ C

Young Modulus (MPa)

30◦ C
102
20◦ C
10◦ C

101

100

10−1
10−8 10−6 10−4 10−2 100 102
Strain rate (s−1 )

Figure 3.13: The initial modulus measured on the first 5% strain on the different
curves for temperatures ranging from 10 ◦C to 70 ◦C. Master curve built at 20 ◦C.

Relaxation experiment

Shift of the glass transition towards lower temperature can be confirmed by the
characteristic times found in relaxation experiments. In these tests, the interlayer
is stretched up to a certain strain and maintained to this strain value for a certain
time. The force is monitored during the loading and constant strain plateau. In
Figure 3.14, the relaxation of the stress is plotted as a function of time when the
strain is maintained at 50% .This experiment was conducted at 20 ◦C. The loading
strain rate was 1 s−1 and the strain was maintained for 1000 s. The stress relaxes
very rapidly in less than 10 s.
Characteristic times of the material behavior can be extracted from these results
using a Prony series model. These characteristic times are compared to the times
found in small strain rheology in Figure 3.15 where the storage modulus calculated
with a Prony series model, is plotted as a function of frequency. The reference
temperature is 20 ◦C. The small strain results (in blue) correspond to the character-
istic times and moduli found in DMA and rheometer tests (Figure 3.11). The large
strain results (in red) corresponds to the previous relaxation experiments (Figure
3.14). At this reference temperature of 20 ◦C, for large strain the glass transition
occurs around 10 Hz whereas for small strain the glass transition occurs around
0.1 Hz. This is similar to the previous result as it implies that the application of
large strain on the samples shifts the glass transition temperature towards shorter
times or lower temperature. This result means that, in order to model the large

43
 10
max = 50%

8
True Stress (MPa)

0
0 20 40
Time (s)
(a) Relaxation during the first 50 s

10
max = 50%

8
True Stress (MPa)

0
0 500 1000
Time (s)
(b) Long term relaxation

Figure 3.14: Relaxation experiments conducted at 20 ◦C at two different strain level

(20 and 50%). Loading strain rate is 1 s−1 and relaxation lasts 1000 s.

44
strain time/temperature dependence of the interlayer, it would be better to rely on
the characteristic times found in the large strain relaxation experiments rather than
on the ones found in the small strain DMA and rheometer tests.

Storage Modulus (MPa)

102

101

Small strain
Large strain
100
10−4 10−2 100 102 104
Frequency (Hz)

Figure 3.15: Prony series using the characteristic times and moduli found in the
relaxation experiment at large strain and in the small strain DMA/Rheometer tests.
The reference temperature is 20 ◦C in both cases. It illustrates the difference induced
by the large strain on the time/temperature superimposition.

Dissipated energy: partial conclusion

One of the main targets of this study of the PVB rheology is to determine the
dissipated energy during polymer stretching (Figure 3.16). A cyclic uniaxial tension
was performed. The area in between the loading and the unloading curve gives the
dissipated energy, while the area under the unloading curve gives an approximation
of the elastically stored energy that is recovered as the polymer is unloaded. Indeed,
during unloading, the thin film of PVB starts to buckle around 100% deformation.
Thus the unloading curve has little meaning below this strain. The polymer will
recover from this viscoelastic deformation after 1 or 2 minutes.
One can calculate the ratio of dissipated energy over total work provided to
stretch the interlayer to 200% strain. This ratio will be denoted Rd . This dissipated
energy ratio as a function of strain rate and temperature is given in Figure 3.16.
The previous different behaviors are recovered in this graph. At low temperature
and high strain rates, the amount of dissipated energy is higher than 80%. This
result is similar to the one found by Nourry during impact experiments on lami-
nated glass. He found indeed that about 87% of the impact energy was dissipated,

45
 80

True Stress (MPa)

60
Dissipated energy
40

20
Elastic energy

0
0 0.5 1 1.5 2
Nominal Strain

Figure 3.16: Dissipated energy during a traction experiment corresponds to the area
in between the loading curve and the unloading curves. The elastic stored energy
corresponds to the area under the unloading curve. Here an example on a uniaxial
tension test at 20 ◦C and 1 s−1 .

mainly by interlayer deformation, during the impact. This huge amount of dissi-
pated energy is due to the glassy/plastic nature of the PBV. The yield stress and
the stiff unloading slope observed in the previous cyclic experiments (Figure 3.12(a))
are indeed similar to the one observed in a plastic material. Other interlayers used
for laminated glass such as the SentryGlass ionomer from Dupont, presents similar
plastic behavior which lead to large amount of dissipated energy. However, PVB
interlayers present also viscoelastic properties which enable full strain recovery and
no permanent deformation even at low temperature and high strain rates.
The ratio of dissipated energy strongly decreases as the temperature increases
above the glass transition temperature from 30 ◦C to 50 ◦C, the polymer gets closer
to an elastic rubber. At higher temperature (70 ◦C), the polymer enters an other
dissipation area. The dissipation ratio increases a little bit compared to the one
observed for 50 ◦C. The interlayer behavior, could be assimilated to a viscoelastic
fluid in this domain of strain rates and temperatures.

3.4 Strain induced birefringence

Birefringent experiments are used to understand the relationship between the inter-
layer behavior and its chemical structure. Indeed, the photoelastic signal is related
to the local polymer chain orientation which is in some way, related to the mechan-
ical loading. These results will lead to propose a rheological model for the material.

46
 100

Dissipated Energy Ratio (%)

Glassy/Plastic
80

60 Viscous
Rubber
40
˙ = 1s−1 Viscous
20 ˙ = 0.1s−1 Fluid
˙ = 0.01s−1
˙ = 0.001s−1
0
10 20 30 50 70
T ( C)
◦

Figure 3.17: Ratio of the elastically stored energy over the dissipated energy as
a function of the temperature and loading speed. Different dissipation regimes
(Plastic, viscous rubbery and viscous fluid) can be distinguished.

3.4.1 Influence of strain rate and temperature on birefrin-

gence
Some photoelastic experiments (see 2.6.3) were performed on the PVB interlayer
during uniaxial tension at controlled strain rate and temperature. The experiments
were performed with a red filtered light. A typical intensity signal observed during
this experiment at 20 ◦C and 1 s−1 is plotted in Figure 3.18. The intensity is a mean
value, measured on a square of approximately 10px in the middle of the stretched
sample (red square on the frames in Figure 3.18(b)). The intensity oscillates between
bright and dark birefringence fringes. When the intensity is maximal the delay
introduced by the material in the light propagation results in a polarized wave which
is in the same direction as the polarizer (after the quarter wave length plate). The
fringe order N is the number of time the received light gets to maximal intensity.
The initial intensity being maximal, it corresponds to a fringe order of N = 0. In
this experiment, the interlayer is stretched up to 100% strain and there are 6 fringes
the last one being of N = 5th order.
The fringe order also depends on time and temperature (Figure 3.19 and 3.20).
At 20 ◦C, three different strain rates were tested 0.01 s−1 , 0.1 s−1 and 1 s−1 in a
uniaxial tension test up to 200%. The result is that the faster the strain the higher
the fringe order. Similarly, at 0.1 s−1 three temperatures were tested 20 ◦C, 50 ◦C and
70 ◦C in a uniaxial tension test up to 200%. We find that the lower the temperature,

47
 120

100

80
Intensity

60

40

20

0
0 0.5 1
Strain
(a) Birefringence normalized intensity

(b) Birefringence experiment

Figure 3.18: Typical birefringence signal during a 20 ◦C and 1 s−1 uniaxial tension
test on the PVB interlayer stretched up to 100%.

48
the higher the fringe order.

Birefringence Order
6

2 ˙ = 1s−1
˙ = 0.1s−1
˙ = 0.01s−1
0
0 1 2 3
Strain

Figure 3.19: Evolution of the birefringence order with strain (up to 200%) as a
function of the applied strain rate at 20 ◦C. Three velocities were compared 0.01 s−1 ,
0.1 s−1 and 1 s−1 .

Moreover, the birefringence is not directly proportional to strain. Considering

the Gaussian network as described by Treloar [17], the fringe order is supposed to
be proportional to the difference of principal stresses (Equation 2.6.3). However,
the non linearities introduced by this simple theory of the polymer network are
not sufficient to describe the fringes observed here. Indeed, the same results as in
Figure 3.19 are plotted in Figure 3.21, as a function of the Gaussian theory principal
stress difference (for an incompressible material in uniaxial tension) and with a
corrected value of the interlayer thickness and the relationship diverges from the
simple proportionality (that would have been expected) especially at larger strains.
It is also interesting to notice that, after a uniaxial tension test on the interlayer
up to 200% deformation, there is no permanent deformation when the sample is
removed from the clamps at 20 ◦C whereas at 50 ◦C and 70 ◦C, the remnant defor-
mation described earlier is clearly visible (Figure 3.22).

3.4.2 Birefringence during relaxation experiment

An interesting result arises from a photoelastic measurement during a relaxation
test on the PVB interlayer. The interlayer is stretched at 20 ◦C at 0.1 s−1 up to
200%. The polymer is then maintained at this level of deformation during 1000 s.
In figure 3.23, during the loading phase, the usual fringe pattern with 8 fringes (up

49
 8

Birefringence Order 6

2 T = 20◦ C
T = 50◦ C
T = 70◦ C
0
0 1 2 3 4
Strain

Figure 3.20: Evolution of the birefringence order with strain (up to 200%) as a
function of the temperature at 0.1 s−1 applied strain rate. Three temperatures were
compared 20 s−1 , 50 s−1 and 70 s−1 .

8
Birefringence Order

2 ˙ = 1s−1
˙ = 0.1s−1
˙ = 0.01s−1
0
0 1 2 3 4
2
(λ − 1/λ).t

Figure 3.21: Evolution of the birefringence order with Gaussian strain corrected
by the calculated thickness t (incompressible material) as a function of the applied
strain rate at 20 ◦C. Three velocities were compared 0.01 s−1 , 0.1 s−1 and 1 s−1 . λ is
the stretch applied onto the interlayer.

50
Figure 3.22: Four different samples initially with the same 20 mm length. From left
to right: Reference sample in the initial shape, sample stretched at 200% at 20 ◦C,
sample stretched at 200% at 50 ◦C and sample stretched at 200% at 70 ◦C. The
samples are only clamped by their upper end despite what can be seen from the
picture.

to fringe order N = 7) is observed. However, during, the relaxation phase, one can
see that while the stress relaxes from 180 MPa to 25 MPa in 1000 s (with a very fast
decrease down to 80 MPa in the first 10 s), the birefringence signal only varies from
half a fringe. The photoelastic signal seems to be quenched by the constant strain
imposed on the material.
200
Intensity
True Stress

40
150
True Stress (MPa)
Intensity

100

20

50

0 0
10−1 100 101 102 103
Time (s)

Figure 3.23: Relaxation experiment followed by photoelasticity. Strain increases

up to 200% at 0.1 s−1 and remains constant for 1000 s. The stress relaxes but the
photoelastic signal is not recovered.

The same experiment (loading and relaxation) was made at the same tempera-
ture 20 ◦C and at the same strain rate 0.01 s−1 . However, this time the temperature

51
was increased after 500 s during the relaxation phase (Figure 3.24). In a first ex-
periment the temperature was increased to 50 ◦C. In a second experiment, the
temperature was increased to 70 ◦C. In both case, the result was completely dif-
ferent from the reference experiment in which the temperature remained constant
at 20 ◦C. During the loading phase, the fringe order goes to N = 7. Only half a
fringe is recovered at 20 ◦C. The increase of temperature at 50 ◦C leads to a partial
recovery of the fringe order as four fringes are passing in the reverse direction. At
70 ◦C, the recovery is more complete as 6 fringes are passing in the reverse direction.
Thus the increase in temperature seems to unblock the birefringence mechanisms
which can relax at constant macroscopic strain.

1.2
Relative Intensity

1
1.2
0.8
Relative Intensity

0.8 0.6
0.6 0.4
0.4 Heating Reference 20◦ C
Reference 20◦ C 0.2 50 ◦ C
0.2 50 ◦ C 70 ◦ C
70 ◦ C 0
0 50 500 1000 1500
0 5 10 15 20
Time (s) Time (s)
(a) Loading (b) Relaxation

Figure 3.24: Relaxation experiment while heating. A reference at 20 ◦C is displayed.

Two temperatures were tested 50 ◦C and 70 ◦C. Loading rate was 0.1 s−1 (a). Re-
laxation lasts 1000 s. Heating start at 500 s after the end of the loading.

3.4.3 Partial conclusion

It is now important to remember that the PVB interlayer is not a crosslinked poly-
mer. The time and temperature dependence of the photoelastic signal can be ex-
plained by the viscoelastic properties of an entangled network. However, at constant
strain, the stress relaxation observed should also correspond to a simultaneous relax-
ation of the polymer chain alignment. Indeed, the entangled network which suppors
the stress increase during the fast loading should, during the long period of constant
strain, relax to its initial randomly organized state. The fact that the photoelastic
signal does not evolve significantly a lot during the relaxation of the stress might
indicate the existence of distinct phase or of two networks in the material. The fol-

52
lowing part of this chapter will focus on presenting some evidence of such a second
phase. The existence of such a second phase in the PVB has been suggested, from
less complete experiments, by Mertz [18] and Schaefer [23].

3.5 Evidence of a second phase

In order to show the existence of a second phase in the material, DSC experiments
were conducted on the PVB. Indeed, Mertz [18] in similar experiments found the
trace of a second phase. This phase should be more organized than the rest of the
polymer network. The DSC results will be confirmed by SAXS experiments made
on the PVB during stretching. Finally, from these experimental observations a first
picture of the material will arise.

3.5.1 An exothermic signal

During a DSC experiment on the PVB interlayer, two important variations can be
observed in the heat capacity. Around 20 ◦C a faster decrease of the heat capac-
ity corresponds to the glass transition of the polymer. Around 70 ◦C (from 50 ◦C
to 100 ◦C) a fusion-like signal is observed which does not correspond to the glass
transition as suggested in [24]. After immediate cooling (red curve), a second tem-
perature ramp at 20 ◦C min−1 does not display the same fusion-like signal around
70 ◦C. However, if during the cooling, a pause of 9 h at 40 ◦C is observed, the fol-
lowing temperature ramp displays a partial recovery of the fusion-like signal. Thus,
the polymer structure leading to this fusion-like signal is not permanently erased
but recovers slowly at ambient temperature. This experiment and the observations
are identical to the one made by Mertz in [18].
The fusion like signal also suggests some form of order to disorder transition.
Thus, there might be some regions in the polymer that present a higher level of
organization than the rest of the polymer matrix. If this hypothesis is correct, these
regions should be visible in other experiment such as Small Angle X-ray Scattering
experiments.

3.5.2 Evidence through X-ray scattering

SAXS experiment were performed on the PVB interlayer at two different tempera-
tures (23 ◦C and 70 ◦C) during uniaxial tension test at a strain rate of 0.01 s−1 up to
300% deformation.

53
 0.5
1st cycle
2nd cycle
0 3rd cycle 9h 40◦ C
1st cycle
-0.28
Cp (mJ/K)

2nd cycle
3rd cycle 9h 40◦ C
-0.5 -0.3

Cp (mJ/K)
-0.32
-1
-0.34

-1.5 -0.36
-50 0 50 100 150 40 60 80 100
Temperature (◦ C) Temperature (◦ C)
(a) The three heating ramps (b) Zoom

Figure 3.25: Three DSC cycles at 20 ◦C min−1 . First cycle shows an exothermic -
fusion like signal around 70 ◦C. The second cycle after a fast cooling at 20 ◦C min−1
do not display this signal. Finally the third cycle done after 9 h at 40 ◦C presents a
partial recovery of the signal.

At 23 ◦C in Figure 3.26, there is no change in the SAXS signal in the traction

direction (Figure 3.26(a)). In the direction perpendicular to the traction direction
however, a signal enters from the small angle while the strain increases (Figure
3.27(b)). This signal is related to the distance between the more organized regions.
At 23 ◦C when the strain increases, this distance decreases in the direction perpen-
dicular to the traction whereas it increases in the traction direction. The distance
at 100% strain between these objects is small enough to be detected in SAXS ex-
periment. When the strain is higher than 100%, this distance becomes smaller than
d> 2π
10−2
≈300 nm.

Strain Strain
10−2 0 0
0.31 10−6 0.31
0.63 0.63
Intensity

Intensity

0.96 0.96
10−4 1.3 1.3
1.6 10−8 1.6
2 2
2.3 2.3
2.6 2.6
10−6 3 10−10 3

10−2 10−1 100 10−2 10−1 100

q (nm ) −1
q (nm ) −1

(a) Traction direction (b) Direction perpendicular to traction

Figure 3.26: SAXS experiment at 23 ◦C during a uniaxial tension experiment.

54
 On the contrary at 70 ◦C, no signal is visible whatever the direction is (Figure
3.27). This could mean that the distance in between the organized regions is too
large to be seen in the SAXS experiment or more probably as suggested by the DSC
experiment, that these regions have been partially or totally destroyed.

Strain Strain
10−2 0.0022 0.0022
0.24 10−6 0.24
0.54 0.54
Intensity

Intensity
0.84 0.84
10−4 1.1 1.1
1.5 10−8 1.5
1.8 1.8
2.1 2.1
2.4 2.4
10−6 2.7 10−10 2.7

10−2 10−1 100 10−2 10−1 100

q (nm−1 ) q (nm−1 )
(a) Traction direction (b) Direction perpendicular to traction

Figure 3.27: SAXS experiment at 70 ◦C during a uniaxial tension experiment.

3.5.3 A schematic model of the structure

From the DSC and SAXS experiments, a model picture of the PVB chain organi-
zation can be drawn. The PVB used in our experiment contains about 18wt% of
hydroxyl groups and almost 20wt% of plasticizer.
The DSC results suggest that there is a more organized phase in the material bulk
which might loose its organization above 70 ◦C. This phase recovers its organization
after several hours at a temperature of 40 ◦C. At this temperature the molecular
motion is sufficient for the region to reorganize. Klock suggested that this second
phase, is made out of regions of the polymer chains with higher hydrogen bond
concentration [19].
The SAXS experiments show that these more organized regions are at a distance
larger than 300 nm from each other. Similar results were obtained by Schaefer
[23]. In this paper, Schaefer used NMR and SANS experiments to show that the
material can be viewed as a two phase material: 1. A matrix of diluted polymer chain
with a high concentration of plasticizer. 2. Stiffer nodules with a lower plasticizer
concentration. According to the author, these nodules are based on strong hydrogen
bond interaction between hydroxyl groups of different polymer chains. According
to his results, these nodules are about 10 nm to 50 nm large.

55
 Both Mertz and Schaefer point out that more plasticizer results in a higher
segregation of these two phases in the PVB structure.
From these results and from the literature, the image presented in Figure 3.28
can be made. The polymer chains are not chemically crosslinked. However, they
are entangled and physically connected by hydrogen bonds. These physical bonds
can be made between the hydroxyl groups of the same polymer chain (intra chain
H bonds) or with another chain (inter chain H bonds). Moreover, the hydrogen
bonds tend to form strong nodules in which the chains are connected to one another
through multiple hydrogen bonds. The polymer chain connected to the nodules (in
red) are constrained by the other chains in the vicinity of the hydrogen bond cluster.
Thus, these regions are more organized and stiffer than the surrounding matrix. In
this matrix, the chains which are not involved in the nodules (in green), constitute
a softer phase. The plasticizer increases the chain movement in the soft matrix.
The PVB structure can be seen as two mixed chain networks: the first one is not
connected directly to the cluster while the second one is physically crosslinked by
the hydrogen bond nodules.

The existence of these two networks and of the H bond nodules, in the material,
should be related to its rheology. In the final part of this chapter we will try to
derive a rheological model from the PVB structure.

3.6 A rheological model: two dissipation mecha-

nisms
We have seen here that the PVB interlayer presents a time/temperature dependent
behavior. This behavior can be described by a generalized Maxwell model at small
strain. At larger strains, this model no longer describes the PVB behavior, which
presents different types of dissipation mechanisms depending on strain rate and
temperature. At temperatures lower than the glass transition temperature 30 ◦C
or at high strain rate, the material is glassy and plastic. At temperature above
the glass transition temperature, the material behaves as a viscoelastic rubber. At
temperature higher than 70 ◦C, the behavior changes into a highly viscous fluid.
We are going to provide a rheological model that captures these different regimes
and is compatible with the different experimental observations in the photoelastic
measurement and in the DSC and SAXS experiments.

First we will consider the polymer chains alone. Their response can be described

56
 Intra chain H bond

>300nm

Inter chain H bond

Stiffer H bond nodules

Plasticizer molecules
Chains involved in a H bond nodule
Chains free of H bond nodules
Vinylalcohol monomers bearing hydroxyl groups
H bonds

Figure 3.28: A schematic view of the two phases in the PVB. A soft PVB matrix
surrounds stiffer and more organized H bonds nodules. These nodules are at a
distance greater than 300 nm from each other.

57
by a Maxwell spring/dashpot model modified to capture the large deformation (Fig-
ure 3.29). The spring in this model will not present a linear elastic response but
a non linear response such as the one described in the Arruda-Boyce model (See
2.4.2). The long term response corresponds of a spring with a shear modulus µ∞
and a limit chain extension λm . The other characteristic times τi are modeled by
a dashpot with a similar Arruda-Boyce spring attached (µi ). Here for the sake of
simplicity only one characteristic time is represented.

µi
µ∞
τi

Figure 3.29: PVB chain rheological model.

This model will properly describe a time dependent non linear material. How-
ever, this model is not sufficient to explain some results which have been previously
described.
First at high strain rate and low temperature, a plastic like behavior has been
observed. The dissipated energy does not decrease as the temperature decreases.
On the contrary, in this model, as soon as the material is far enough below the glass
transition, it will recover an elastic behavior with little dissipation.
Secondly, the birefringence order increases during a traction experiment but it
does not change significantly during a relaxation experiment. The birefringence
order is also higher at a given strain for faster strain rates and lower temperature.
In the model, if the birefringent spring is associated to a short time dashpot, the
signal will be more important at high strain rate and low temperature but it will
relax rapidly with the stress. On the contrary, the birefringent signal could be
associated to a long term dashpot, in order to block it during the relaxation. In
this case the birefringent signal will not be more important at low temperature and
high strain rates than at high temperature and low strain rates, which again is in
contradiction with some of the data.

In order to properly model the birefringent behavior along with the plastic like
behavior at low temperature and high strain rate, an other branch has to be added in
parallel of the previous model (Figure 3.30). This second branch will be responsible
for most of the photoelastic signal. The time dependence and the elasticity related
to this branch are, similarly to the previous branches, described by a dashpot with

58
a characteristic time τγ and a non linear Arruda-Boyce spring with a modulus µ0γ .
In serie of this dashpot and this spring, another block enables the increase of the
photoelastic signal at short time and its conservation at constant strain level. This
block is constituted of a slider with a threshold stress σγ and a spring with a modulus
µγ . The slider is only activated if the stress increases above a threshold stress. Thus
during the loading the birefringence signal changes. During the relaxation test, the
slider will be blocked and the birefringence signal in the spring µγ will not change
while the stress stored in the spring µ0γ is free to relax.

σγ µγ
µ0γ
τγ

Figure 3.30: Plastic part of the rheological model of the PVB

The resulting final model is represented in Figure 3.31. This rheological model
will also explain the different dissipation regimes observed in the large strain exper-
iments. At temperatures above the glass transition temperature, the plastic part
of the model will not be triggered because the stress remains below threshold; the
material will be a viscoelastic rubber. At lower temperature or higher strain rate,
this second plastic branch will be activated when the stress reaches the critical σγ
leading to plastic dissipation. This model is close to the Bergström-Boyce model
[25] in which two parallel mechanisms are used to describe the mechanical behavior
of the interlayer. A first channel is hyper-elastic while the second one introduces
dissipation in a purely plastic manner. The model has been extended with the intro-
duction of a second viscoelastic dissipation mechanism in the hyper-elastic branch
of the model [26].
Finally, PVB presents hydroxyl groups which are able to form stiffer nodules
and we have seen that two networks can be used to describe the polymer structure.
The two dissipative branches in the rheological model could also be understood as
representative of these two networks. The chains which are not involved in the
nodules are modeled by the visco-hyperelastic part of the model whereas the chains
involved in the H bond nodules are modeled by the plastic part of the model.

59
 σγ µγ µi
µ∞
µ0γ
τi
τγ

Figure 3.31: A rheological model for the interlayer. Two dissipation mechanisms are
present in this model.

Main results
• Rheological study in small and large strain regimes.

• High level of dissipation at large strain.

• Birefringent behavior:

– Photoelastic signal dependent on the strain rate and tempera-

ture.
– Relaxation of stress is not correlated to a recovery of the photoe-
lastic signal.

• PVB structure:

– H bonds nodules: stiffer and more organized.

– Two networks: Chain involved in nodules vs. Chain free of nod-
ules.

• A rheological model: two branches of dissipation (plastic and visco-

hyperelastic).

60
Chapter 4

Model delamination experiment

4.1 Introduction
The previous chapter 3, has shown the origin of the dissipation of energy in the
polymer structure. The model experiment will now be used to explore the loading
of the interlayer during its delamination from glass. The rheology of the PVB
interlayer in uniaxial tension displays different dissipation mechanisms. They are
activated at different strain rates and temperatures. In an impact test, the interlayer
is subjected to different mechanical loadings mainly uniaxial tension and shear.
It might also encounter different strain rates regimes as it delaminates from the
glass. Our purpose is to identify the different zones of stretching and strain rates
encountered by the interlayer in these zones. This will enable a direct link between
the rheology of the PVB and the calculation of the energy dissipated during the
delamination process. Due to the extreme complexity of an impact test, we decided
to introduce a model experiment, which focuses only on a single type of mechanical
loading occurring during the impact. Different geometries have been used in previous
studies to characterize the interlayer/glass delamination mechanisms.
A peel test can be used to study the glass/interlayer adhesion. Pelfrene [27]
used such a peel experiment. A rigid back is attached to the interlayer to prevent
the stretching of the free arm during the delamination. In this test, the peeling
force was measured and used to build a finite element model of the experiment.
Using cohesive elements to represent the interface and a linear viscoelastic model
for the interlayer rheology, they showed that most of the energy is dissipated by the
interlayer deformation in the close vicinity of the crack tip. The linear viscoelastic
modeling of the interlayer in this region was found to be no longer appropriate due
to the large deformation involved (300%).

61
 In [28], Seshadri presented a different testing geometry. The compressive shear
strength (CSS) test consists in a shear and compression combined loading on a
laminated glass sample (Figure 4.1). The laminated sample is at 45◦ from the
loading direction. The polymer is confined in between the two glass plates and is
free to deform in between. This loading might be more representative of the shear
applied on the interlayer. Thanks to this test and a finite element model, they were
able to extract a value for the apparent fracture energy.

Figure 4.1: Schematic view of the CSS test in which the interlayer in between the
two glass plates is subjected to both shear and compression.

However, both these tests did not describe the bridging of the glass parts by the
stretched interlayer described by Seshadri in [12] and [29]. In order to do so Seshadri
used the Through Crack Tensile Test (Figure 4.2). It consists of a uniaxial tension
applied on a pre-cracked laminated glass sample. This experiment has been also used
in other studies on laminated glass ([30], [31], [32], [33], [34]). This geometry focuses
on the second phase of the impact (Figure 1.3) during which the plate goes into a
macroscopic bending. During this step, the glass parts still stick to the interlayer
and are going away from each other. This leads to the delamination of the interlayer
and its stretching in between the delamination fronts. Due to the large curvature of
the flexed structure, the local mechanical loading in between the glass parts can be
assimilated to a uniaxial tension. This model test reproduces a mode I stretching
of the delaminated interlayer. Apart from Delince [33], all these studies found that
there can be a stable delamination growth. In this regime, the polymer is stretched
to a high level of deformation leading to a large amount of dissipated energy.

In this chapter, we will show that in the TCT test, there is a limited range

62
 F δ

Delamination
Front
a

a
xF

Figure 4.2: Through Crack Tensile Test (TCT Test), model experiment. A uniaxial
tension test is conducted on a pre-cracked laminated glass sample. The interlayer is
stretched in between the delamination fronts which propagate in opposite directions.

of temperature and applied velocity in which the delamination is steady and the
delamination fronts straight. We will show, that inside this stable delamination
regime, the polymer stretching from 0% strain to a high strain value, occurs on a
characteristic length scale close to the interlayer thickness.

4.2 Description of a typical Through Crack Ten-

sile Test
The TCT Test has been described in 2.5.2. Experiments were conducted at various
temperatures 5 ◦C, 20 ◦C and 50 ◦C and with different applied velocities ranging from
0.03 mm s−1 to 100 mm s−1 .
During a typical TCT Test at 20 ◦C and 10 mm s−1 four delamination fronts
propagate at a constant pace in two opposite directions. In between these fronts the
interlayer is stretched (Figure 4.3).
As found by [30], [12] and [31], a steady state delamination is observed. The
delamination force FTCT reaches a plateau which mean value FTCT
0
is approximately
500 N after a first overshoot (Figure 4.4).
During the experiment, the upper glass edge velocity δ̇ is controlled. The position
of the fronts relative to the glass edge a is recorded. From this measurement, the
stretch (λTCT ) can be determined as the ratio of the delaminated interlayer length
(δ + 2a) over the initial length occupied by this part of the delaminated interlayer

63
 Upper clamp

δ
50 mm
xF a

0s 1s 2s Time
Glass cracks Delamination
fronts

Figure 4.3: A typical TCT experiment at 20 ◦C and 10 mm s−1 . Sample width is

50 mm. δ is the upper clamp/upper glass displacement. a will denote the position
of the delamination front relative to the glass edge. xf will be the distance of a
given point to the point at the vertical of the delamination front.

600

500
Force FTCT (N)

400

300

200

100

0
0 1 2 3 4
Time (s)

Figure 4.4: Delamination forces FTCT reaches a steady state which mean value is
0
FTCT = 500 N during a TCT experiment at 20 ◦C and 10 mm s−1 .

64
(2a) which is the length between the delamination fronts and the glass edge (Figure
4.2): λTCT = δ+2a
2a
. The nominal strain is then defined as TCT = λTCT − 1 = δ
2a
. In
Figure 4.5, the upper glass displacement δ is plotted as a function of the delaminated
length 2a. As the fronts propagate symmetrically the slope of the curve gives us
a mean value of the strain 0TCT close to 1.4. As shown on Figure 4.6, the strain
is also constant through the whole test. This underlines the fact that the front
propagation occurs at a velocity close to 3.5 mm s−1 which is proportional to the
applied displacement velocity (δ̇ = 2ȧ).

40

30
δ (mm)

20
0TCT ≈ 1.4
10

0
0 10 20 30
2a (mm)

Figure 4.5: In the 20 ◦C and 10 mm s−1 TCT Test, the symmetric delamination
of the fronts occurs at a constant velocity close to 3.5 mm s−1 proportional to the
applied displacement velocity. Thus the delamination length 2a is proportional to
the displacement applied on the glass δ.

4.3 Influence of velocity and temperature on de-

lamination: phase diagram

4.3.1 Results
The steady state delamination described in the previous part, is not observed for
all temperatures and velocities. At a given temperature, there is an intermediate
velocity regime in which the delamination occurs in a symmetric and steady manner
as described before. At lower imposed velocities, the delamination fronts are no
longer straight and start to undulate. Propagation often stops after some time

65
 2.5

Strain TCT 1.5

0.5

0
0 1 2 3 4
Time (s)

Figure 4.6: In this 20 ◦C and 10 mm s−1 TCT Test, the nominal strain TCT reaches
a steady state plateau which mean value is 0TCT = 1.4 and that is related to the
symmetric delamination of the fronts at a constant velocity.

leading to the rupture of the interlayer (Figure 4.7(a)). Finally at higher pace, even
if the delamination occurs as in the steady state delamination regime, the interlayer
toughness is not sufficient which often leads to a brittle fracture of the interlayer
(Figure 4.7(a)). It is worth noticing that the brittle rupture of the interlayer in the
high velocity regime is different from the one in the low velocity regime, in which
the crack tip is largely blunted. Measurements could only be made in the steady
state regime as it was not possible to obtain a stable delamination in the lower or
higher velocity regimes.
In Figure 4.8, the mean delamination force FTCT
0
is plotted for all the displace-
ment rates and temperatures investigated. There are strong variations of the force
which increases as the velocity increases or as the temperature decreases. On the
same figure, the delamination behavior is divided in three regimes. The limits dis-
played here are roughly estimated.
In Figure 4.9, the average nominal strain 0TCT is presented as a function of
applied velocity and temperature. There is a significant dependence of the strain
with the temperature. Indeed at 5 ◦C, the strain is in a range from 1.05 to 1.3, at
20 ◦C it ranges from 1.1 to 1.6 and at 50 ◦C the strain goes from 1.8 to 2.05 for the
different velocities. In the steady state delamination regime, the strain is slightly
increasing with the velocity at 5 ◦C. This nominal strain increase is more important
at 20 ◦C. At 50 ◦C the strain is almost constant with the applied velocity. The large
error bars observed at 20 ◦C for the highest velocities correspond to cases where the

66
 (a) TCT test at 20 ◦C and 0.01 mm s−1 in the low velocities/high temperatures
regime.

(b) TCT test at 20 ◦C and 10 mm s−1 in the steady state regime.

(c) TCT test at 5 ◦C and 10 mm s−1 at the limit of the high velocities/low
temperatures regime.

Figure 4.7: Out of the steady state delamination regime, two different behaviors
are observed. (a) At low velocities or high temperatures, the delamination fronts
undulate and stop sometimes leading to the interlayer fracture. (b) In the steady
state regime. (c) At high velocities or low temperatures, the delamination propagates
symmetrically but the interlayer breaks rapidly in a brittle manner.

67
 1000
5◦ C
20◦ C
50◦ C
800

Force FTCT (N) 600

Steady State Brittle
0

400 Delamination Regime

200
Unstable
Delamination
0
10−2 10−1 100 101 102
Velocity δ̇ (mm/s)

Figure 4.8: The mean steady state delamination force FTCT0

increases with velocity
δ̇ and decreases with temperature. The dark lines arbitrary delimit the three delam-
ination regimes. Measurements were only possible in the intermediate regime where
no buckling of the front (low velocities and high temperatures) or brittle nature of
the interlayer (high velocities and low temperatures) are observed.

interlayer ruptured after a small delamination.

4.3.2 Comparison with previous studies

It has been pointed out in previous studies such as the one conducted by Butchart
[35] or Delincé [33] that the stable delamination regime is not trivial to obtain.
Even if the PVB grades used in these studies are, in some cases, different from the
one we use here, this might reflect the fact that their experiments were conducted
at low imposed velocities and ambient temperature. They observed, as we did,
the same unpredictable behavior of the delamination fronts which stop after a small
delamination length leading to the rupture of the interlayer. Our observations might
give a key to determine if the delamination in the TCT test is going to occur in a
steady state manner.
Furthermore, it appears that the temperature control is critical, as the PVB
glass transition is broad and spans from 20 ◦C to 30 ◦C. Thus around the room
temperature a small variation of the temperature might lead to a dramatic change
in the interlayer behavior leading to a large shift in the stable delamination regime.
This aspect of the problem is not tackled in all the previous studies.
Faster TCT experiments were done by Ferretti in [36]. They reach a stable
delamination regime and conclude, as we do, that the applied velocity is related

68
 2.5

Nominal Strain 0TCT

1.5

0.5
50◦ C
20◦ C
5◦ C
0
10−2 10−1 100 101 102
Velocity δ̇ (mm/s)

Figure 4.9: Dependence of the mean nominal strain 0TCT with velocity and temper-
ature. The dependence with the applied velocities is blurred by large error bars.
This large error in the strain measurement at 20 ◦C for the highest strain rates is
due to early rupture of the interlayer after a small delamination and to the small
amount of data that results of a low video frame rate.

to the delamination stability. Furthermore, they add that this stability is related
to the more important stiffness of the interlayer at higher strain rates. It could be
related to the instabilities we observed at lower strain rates. An hypothesis could
be that, in a softer interlayer, the stress can not be transmitted uniformly along the
front which might lead to its undulation.

4.4 Distribution of the deformation of the inter-

layer in the TCT test
The point of this chapter is to evaluate how the interlayer is loaded as it delaminates
from the glass during impact. This information is not easily accessible during the
impact test (see 1.2.2). The TCT test was used in the previous measurements to
show that the nominal strain during steady state delamination, reaches a constant
average value. As the interlayer behavior is highly dependent on the strain rate it is
also important to determine the length over which the interlayer is stretched from 0
to the average value 0TCT from which the strain rate involved in this region can be
deduced. That would be essential to understand the dissipating mechanisms which
will be studied in the next chapters.

69
 Upper clamp
Delamination front
+ high strain rate
bulk deformation zone

Intermediate
deformation zone

Steady state
deformation zone

Figure 4.10: A photoelastic experiment conducted on a laminated glass sample

during a TCT experiment at 20 ◦C and 10 mm s−1 . Different deformation zones are
visible on this picture. The strain is homogeneous far from the delamination front.

By taking a macroscopic snapshot, in a photoelastic setup, it is possible to ob-

serve the variations of the strain field (Figure 4.10). The vertical fringes correspond
to edge effects. In these regions, the strain might be spatially not homogeneous but
it does not involve high strain rates as the fronts are propagating in a direction or-
thogonal to the fringes. The different zones on which the interlayer is stretched can
be seen on this picture by looking at the horizontal fringes. Close to the delamina-
tion front, the high fringe concentration might be related to a high strain rate region.
A second region is observed over a length scale similar to the sample width, in which
the strain might evolve more slowly. Finally, far from the delamination front and
from the sample edges, the strain seems to be constant and rather homogeneous. In
order to measure more precisely the length of these stretching zones, digital image
correlation and more quantitative photoelastic experiments were performed.

4.4.1 Deformation zone measured by photoelasticity

The size of the fast stretching zone close to the moving delamination fronts can
been observed in photoelastic measurements performed during the through crack
tensile test. Figure 4.11 shows the results of the TCT test conducted at 20 ◦C and
10 mm s−1 , on three samples with different interlayer thicknesses 0.38 mm, 0.76 mm
and 1.52 mm. The photoelastic signal was measured on a zone of approximately
4 mm length, on a vertical line, in the middle of sample. As delamination occurs in
a steady state, the complete spatial signal was reconstructed by shifting the different
curves obtained at different times, using the front velocity.
The birefringence fringes are all concentrated in a zone with a length close to

70
 Intensity
1
h = 0.38 mm
0.5
0 0.2mm
-5 0 5

Intensity
1
h = 0.76 mm
0.5
0 1mm
-5 0 5
Intensity

1
h = 1.52 mm
0.5
0 1.5mm
-5 0 5
d (mm)

Figure 4.11: Photoelastic measurement conducted close to the delamination front

during a TCT experiment at 20 ◦C and 10 mm s−1 .

the interlayer thickness. Thus, most of the stretch occurs over this length scale.
Furthermore one can see that the number of fringes is increasing when the thickness
of the interlayer increases. However, it is difficult from this experiment alone to tell
if this is due to an increase in the stretch or to the change in thickness. In order
to get a more quantitative measurement of the strain profile in this fast stretching
zone, digital image correlation measurements were performed.

4.4.2 Fast stretching zone measured in DIC

A TCT experiment has been conducted at 10 mm s−1 and 20 ◦C, in the steady state
regime. Two interlayer thicknesses were tested: 0.76 mm and 1.52 mm. The delami-
nation front velocity was found to be almost identical for both thicknesses (3 mm s−1
and 4 mm s−1 for the 0.76 mm and 1.52 mm thickness respectively). Far front the
delamination front, the interlayer is stretched to a maximal value close to 0TCT .
Using digital image correlation, the displacements field is recorded while the
delamination occurs. From these displacements, the interlayer nominal strain is
deduced along a vertical line in the middle of the sample. The strain DIC is nor-
malized by the mean constant strain value 0TCT previously measured. The distance
to a point upright the delamination front position xF (Figure 4.2) is normalized by
the interlayer thickness h. In Figure 4.12, the normalized strain along the central
vertical line is displayed as a function of the normalized delamination front position
for two different thicknesses.

71
 The interlayer deforms from 0% to 60% of 0TCT over a length close to the in-
terlayer thickness (Part A of the curve). From the delamination front velocity, one
can deduce that the average strain rate in this region is approximately 3 s−1 . After
this sharp increase of the strain, the polymer reaches the steady state strain at a
much slower pace with a strain rate being close to 0.5 s−1 . We know from our pre-
vious results that eventually, far from the delamination front, the polymer reaches
a constant strain of the order of the mean value 0TCT = 1.4. In this zone, no more
energy will be dissipated. This region is not easily accessible, with the same spatial
resolution. A zoom out will be required to measure the strain using DIC, far from
the delamination fronts.

1
h = 1.52 mm
h = 0.76 mm
0.8

0.6
DIC /0TCT

0.4
A
0.2

-2 -1 0 1 2 3 4
xF /h

0 xF

Figure 4.12: Local strain measurement through digital image correlation at 20 ◦C

and 10 mm s−1 . Two thicknesses h were tested. xF is the position to the point
upright the delamination front. Two strain rate regimes are identified ahead of the
position uprigth the delamination front xF /h = 0. The first regime spans over a
length close to the interlayer thickness.

72
4.4.3 Dependence of the fast stretching zone length on ap-
plied velocity and temperature
The same DIC experiment can be conducted at different velocities and different
temperatures.
The experiment has first been conducted at 20 ◦C for three different velocities
10 mm s−1 , 30 mm s−1 and 60 mm s−1 . As the imposed velocity increases, the de-
lamination front velocity also increases. Thus we found three different delamination
front velocities 4 mm s−1 , 10 mm s−1 and 19 mm s−1 . The local strain was plotted
as a function of the delamination front distance in Figure 4.13(a). The amplitude
of the strain seems to be identical in each case leading to the same TCT far from
the delamination front. This could mean that the mean constant strain TCT does
not change with the velocity at 20 ◦C unlike what was seen in the previous part 4.9,
where the strain seemed to increase with the applied velocity from 1.2 to 2. The
deformation occurs on the same length, which is approximately the thickness of the
interlayer as described before. This implies that the deformation of the interlayer
occurs at different strain rates. In Figure 4.13(b) the local strain is now plotted as
a function of time. The continuous lines have a slope close to the theoretical values
calculated by ˙ = 0TCT /h.Vf ront . The different strain rates found here are 3 s−1 , 4 s−1
and 5.2 s−1 for the different imposed velocities respectively.

1.2
δ̇= 60 mm/s
1 δ̇= 30 mm/s 1
δ̇= 10 mm/s
0.8
0.8
0.6
0.6



0.4
0.4
0.2
0.2 ˙ = 5.2 s−1
0 ˙ = 4.0 s−1
˙ = 3.0 s−1
0
-5 0 5 -0.1 0 0.1 0.2 0.3
xF /h Time (s)
(a) (b)

Figure 4.13: DIC experiment conducted during a TCT test at 20 ◦C and three dif-
ferent imposed velocities 10 mm s−1 , 30 mm s−1 and 60 mm s−1 . (a) The deformation
length does not depend on the imposed velocity. (b) The strain rate increases as
the imposed velocity is increasing.

The DIC experiment was also performed for different temperatures: 5 ◦C, 20 ◦C
and 50 ◦C at 30 mm s−1 . As the steady state delamination regime is highly affected

73
by the temperature it was hard to find a velocity at which the delamination was
stable for the three temperatures. The following results have to be taken with care
and more experiments might be required to validate this result. In Figure 4.14 the
strain is plotted as a function of the distance to the delamination front. At 50 ◦C,
a transient negative strain is observed in the front neighborhood. This effect has
been observed twice at this temperature on two different specimens. As this region
corresponds to the passage of the dark line of the delamination fronts, it could either
be an artifact from the image correlation process or due to the undulation of the
fronts.
Nevertheless, it can be seen that the deformation length scale is equal to the
interlayer thickness at 20 ◦C and 50 ◦C. At 5 ◦C, this length is closer to twice the
interlayer thickness.

1.5
T = 5 ◦C
T = 20 ◦ C
T = 50 ◦ C
1

0.5


-0.5
-2 -1 0 1 2 3 4
xF /h

Figure 4.14: Local strain measurements at 30 mm s−1 and three different tempera-
tures 5 ◦C, 20 ◦C and 50 ◦C.

4.5 Conclusion
During the delamination of the interlayer from glass, a steady state regime can
be reached in a certain range of temperature and velocity. In this regime, the
delamination force and strain reach a constant value both in time and space far from
the delamination front. At lower temperatures or higher velocities, the interlayer is
more susceptible to failure and displays a brittle behavior. At higher temperatures
or lower velocities the delamination is unstable and the front starts to undulate.

74
 The plateau values reached by both the delamination force and strain are related
to a complex combined effect of the interlayer rheology and of the adhesion. The
increase of the delamination force with the increasing velocity or decreasing tem-
perature is similar to the increase in force observed in the rheology of viscoelastic
polymers. The variations displayed by the interlayer strain are more counter intu-
itive. It can not be directly related to the interlayer behavior. Indeed, the strain
seems to increase with the applied velocity but it also increases with the temperature
which is not a common behavior for polymer materials. The strain level might be
related to a combination of a change in adhesion with the temperature and the vari-
ations of the strain hardening presented by the interlayer at large strains (Section
3.3.2).
Finally, we have seen that the deformation of the interlayer is occurring over a
length which is approximately its thickness. On this length it undergoes a defor-
mation from 0 to 60% of the constant nominal strain value 0TCT . This length does
not depend on the imposed velocity and seems to be stable with the temperature
(though this last result has to be confirmed). Thus, the strain rate can be deter-
mined knowing the delamination front velocity, the mean nominal strain 0TCT (two
parameters that depend on the velocity and the temperature) and the interlayer
thickness. Here we have found strain rates ranging from 3 s−1 to 5.2 s−1 for applied
velocities in between 10 mm s−1 and 60 mm s−1 .
We are now going to use these results to evaluate the dissipated energy during
the delamination.

75
Main results
• Steady state delamination regime.

– At intermediate velocity and temperature: symmetric and steady

delamination
– At high velocity or low temperature: symmetric delamination
but brittle interlayer.
– At low velocity or high temperature: unstable delamination and
fronts undulates.

• The interlayer is stretched from 0% to ≈ 0TCT :

– In a fast stretching zone, which length is close to the interlayer

thickness, the interlayer is stretched to 60% of 0TCT .
– In an intermediate zone, the interlayer is stretched more slowly
to the maximal nominal strain
– The interlayer reaches a constant nominal strain value far from
the delamination front where the strain rate is null

76
Chapter 5

Energy dissipation during

delamination

5.1 Introduction
During the delamination, we have seen that the interlayer deforms over a length
equal to its thickness equal at 0.76 mm. At 20 ◦C, the fronts propagate at high
velocities of about 3 mm s−1 and the interlayer is stretched up to 140% nominal
strain. Consequently, the strain rates that are applied on the polymer are about
3 s−1 . These high strain rates, on a 0.76 mm thick volume, could involve a large
amount of dissipated energy despite the weak hydrogen bonded interface.

In the literature, this large amount of dissipated energy during the propagation
of a crack inside a viscoelastic material or at the interface between a viscoelastic
polymer and a rigid substrate has been a long time puzzling problem. During the
fracture of a viscoelastic material, even if the rupture of chemical or physical bonds
only requires 0.01 J m−2 to 0.1 J m−2 , huge macroscopic fracture energy of the order
of 1 kJ m−2 are generally recorded. Schapery describes the crack tip has a region
in which the interlayer behavior is strongly non-linear and viscoelastic [37]. De
Gennes [38] suggested in his trumpet model that because of the time dependence
of the polymer behavior, there were different dissipation zones ahead of the crack
tip. Later Saulnier et al. in [39] have shown that this simple linear viscoelastic
description of the material can explain the large enhancement of the dissipated
energy observed during the crack propagation. However, Gent [40] used the same
viscoelastic time dependence for the energy release rate to show that it leads to a
dissipation zone at the crack tip with an unrealistic size, close to 1 Å. As explained

77
by Barthel et al. in [41] the coupling between the remote loading and the crack
tip dissipation phenomenon is done through an intermediate viscous/large strain
dissipation zone. According to Creton et al. [42], this could explain both the high
level of dissipation and a larger crack tip process zone. Newby [43] and Amouroux
[44] have shown that for weak interfaces, energy dissipation is low and dominated
by interfacial slippage. A strong enough interface is required to stretch the polymer
and dissipate large amounts of energy.

First, the delamination experiment results will be used to determine the macro-
scopic work of fracture. Then we will see how this work is related to the stretching
of the interlayer. Finally, we will provide a model of the delamination and the
stretching of the interlayer, which will enable a calculation of the dissipated energy.

5.2 Macroscopic work of fracture

The interlayer is stretched to large nominal strains ranging from 100 to 200% during
the delamination. We have seen that a large amount of the work provided to stretch
the interlayer is dissipated (see 3.3.2). The macroscopic work of fracture provided
during the delamination experiment is either stored elastically in the debonded
interlayer or dissipated when this part of the interlayer is stretched. Finally, only a
small amount of energy will reach the delamination front to break the interface.
Using the previous measurements at different temperatures and different applied
velocities, one can first evaluate the macroscopic work of fracture provided in the
TCT test. The total macroscopic work of fracture is defined by:

1 dWTCT 1 dδ 1
Gm,total = = FTCT0 = 2FTCT0 TCT0 (5.2.1)
w da w da w
where WTCT is the work provided to the system during the delamination.
For symmetry reasons we are only interested in half the sample. Thus we will
define the macroscopic work of fracture during the steady state delamination as:

0
FTCT
Gm =  (5.2.2)
w TCT
0

Using the data from Figure 4.8 and Figure 4.9 Gm has been calculated (Fig-
ure 5.1). It is found to increase with increasing loading speed and decrease with
increasing temperature. More specifically, as the velocity increases from 1 mm s−1
to 60 mm s−1 , Gm increases from 7 kJ m−2 to 26 kJ m−2 at 20 ◦C. The increase in

78
temperature from 5 ◦C to 50 ◦C at 10 mm s−1 leads to a dramatic decrease in Gm
from 22 kJ m−2 to 3 kJ m−2 .

Macroscopic Work of Fracture Gm (kJ/m2 )

35
5◦ C
20◦ C
30
50◦ C

25

20

15

10

0
10−2 10−1 100 101 102
Velocity δ̇ (mm/s)

Figure 5.1: Impact of the velocity and temperature on the macroscopic work of
fracture.

5.3 Impact of the interlayer thickness

The steady state value of the nominal strain found in previous experiments (see
section 4.4.1 and following), was shown to be reached rapidly over a length scale
equal to the interlayer thickness. The influence of this thickness on the TCT Test
response was studied. Four different interlayer thicknesses were used: 0.38 mm,
0.76 mm, 1.14 mm and 1.52 mm. A TCT Test was conducted on each of them at
T =20 ◦C and δ̇ =10 mm s−1 in the steady state delamination regime. The increase
in the interlayer thickness leads to a proportional increase in the measured force
(FTCT
0
) from 220 N to 710 N, while the interlayer strain (0TCT ) decreases slightly
from 1.4 to 1.2 (Figure 5.2).
The dissipated energy is thus increasing from 7 kJ m−2 to 16 kJ m−2 as the thick-
ness increases from 0.38 mm to 1.52 mm (Figure 5.3). This increase is linear with a
slope Πbulk equal to 8 MJ m−3 and a non-null intercept 2Γcrack equal to 4 kJ m−2 .
i
This interlayer thickness dependence also reflects that the initial transient regime
during which the force and strain increase, only lasts the time for a length equivalent

79
 1000 2.5
Mean Delamination Force (N)

Nominal Strain 0TCT

800 2

600 1.5

400 1

200 0.5

0 0
0 0.5 1 1.5 0 0.5 1 1.5
Thickness (mm) Thickness h (mm)
(a) Delamination force (b) Delamination Strain

Figure 5.2: Impact on the delamination force and the interlayer stretch of the in-
terlayer thickness. Four thicknesses were tested 0.38 mm, 0.76 mm, 1.14 mm and
1.52 mm.
Macroscopic Work of Fracture Gm (kJ/m2 )

20

15

10
dGm
dh
= Πbulk
= 8±0.9 MJ.m−3
5

2Γcrack = 4±0.8 kJ.m−2

0
0 0.5 1 1.5
Thickness h (mm)

Figure 5.3: Effect of the interlayer thickness on the macroscopic work of fracture.

80
to the interlayer thickness to be delaminated. This can be seen in Figure 5.4 where
the delamination force is plotted as a function of the delamination length normalized
by the interlayer thickness. The steady state starts for the same normalized length
1 for the four thicknesses.
800

600
Force FTCT (N)

400

h = 1.52 mm
200 h = 1.14 mm
h = 0.76 mm
h = 0.38 mm
0
0 1 2 3 4 5
Normalized delamination length a/h

Figure 5.4: The force is plotted as a function of the delamination length normalized
by the interlayer thickness. The transitory regime only last until a length equal to
the interlayer thickness is delaminated.

5.4 Different zones of dissipation

The dissipation processes leading to the large macroscopic work of fracture are not
homogeneous in the interlayer. Gm was found to be an increasing linear function of
the interlayer thickness (Figure 5.3).
In analogy with the method of the Essential Work of Fracture [45], we can
tentatively separate the bulk stretching work due to the deformation of the interlayer
from a near crack work of fracture, which includes the interfacial bond rupture
between the interlayer and the glass and the dissipation due to the stretching of
the interlayer close to the delamination front. The Near crack work of fracture will
depend on the loading mode mixity and on the interlayer confinement, but in the
explored range of interlayer thickness, it does not depend on the thickness (equation
5.4.1). Hence, on Figure 5.3 the slope of the line gives the volumic dissipation density
(Πbulk ) related to the bulk stretching mechanisms: Γbulk = Πbulk h. The intercept
gives near crack work of fracture: Γcrack (per unit area). For half a sample we can
write:

81
 Gm = 2Γcrack + Πbulk h (5.4.1)

After the fast stretching zone, the polymer undergoes some deformation at a
much slower pace. Because of the lower strain rate and of the smaller strain ampli-
tude in this third deformation phase, this contribution to the macroscopic work of
fracture should be much smaller than in the previous zones. Eventually, far from the
delamination front the polymer finally reaches a constant level of nominal strain. In
this last zone, the energy is only stored elastically and its volume density is constant,
no energy will be dissipated.
Hence in the steady state delamination regime we have shown that there is several
dissipation zones (Figure 5.5):

• Close to the delamination front, a near crack dissipation zone. This zone is
experimentally out of our reach because of its small size.

• Ahead of this zone, we have found a fast stretching zone the length of which
is equal to the interlayer thickness. The polymer undergoes a large change of
strain at high strain rate leading to a large amount of dissipated energy.

• An intermediate zone, in which the interlayer is stretched up to the steady state

value of strain at a much slower pace, leading to lower energy dissipation.

• Finally far from the delamination fronts, there is a constant nominal strain
zone where no energy is dissipated.

Γcrack
h Strain w
0 Πbulk
h
Strain rate

Figure 5.5: Far field/Delamination front link and energy dissipation

82
5.5 Modeling the bulk stretching of the interlayer

An important part of the work occurs in the fast stretching zone. In this region, the
interlayer is stretched mainly in uniaxial tension. Thus, a uniaxial tension test at the
appropriate strain rate could be an easy way to estimate the amount of dissipated
energy in the TCT test. This final part of the discussion will focus on modeling the
fast stretching zone.
The volume of polymer involved in the stretching of the interlayer as the front
advances a distance da has to be evaluated in order to estimate the amount of
energy dissipated during the delamination. The remote displacement dδ is applied
with a constant loading velocity δ̇ that corresponds to a constant measured force
0
FTCT while the four fronts will simultaneously move a distance da. As described
before, the delamination fronts are moving at a constant velocity and the following
relationship can be written:

2
da = Vfront dt ≈ dδ (5.5.1)
TCT
where Vfront is the delamination front velocity and dt is a small time increment.
The interlayer in the volume whda will first pass above the delamination fronts
and then will overcome the large deformation process in the fast stretching zone.
Thus, in a steady state regime for all the fronts to advance by da, the total macro-
scopic energy is:

Etot = 2Gm wda = 2 (2Γcrack wda + Πbulk whda) (5.5.2)

It is possible to estimate the volumic density of the bulk stretching work from
a simple uniaxial tension curve. In this zone the polymer is stretched from 0 to a
strain of 0.8 (60% of 1.4 see 4.12) at 3 s−1 . Hence, the bulk stretching work density
is given:

Z fTCT ∗TCT
Πcalculated ≈ σT dT (5.5.3)
0

With σT and T the stress and strain values measured during the uniaxial tension
test. fTCT is the fraction of the constant mean nominal strain value TCT reached
by the interlayer in the fast stretching zone. Here fTCT is about 60% of TCT .
This yields Πcalculated ≈ 9 MJ m−3 which is comparable to the value of 8 MJ m−3
found by our previous measurements in Figure 5.3.

83
5.6 Dissipated energy
The interlayer stretching, in the fast stretching zone, is approximately a uniaxial
tension. The dissipated energy can be evaluated from the macroscopic work of
fracture as if the delamination was a uniaxial tension test. In section 3.3.2, we have
defined a dissipated energy ratio Rd for the uniaxial tension test. This ratio can be
applied to the macroscopic work of fracture to estimate the dissipated energy:

FTCT
Gdiss = Rd ∗ Gm = Rd ∗ TCT (5.6.1)
w
However, this ratio depends both on velocity and temperature as displayed in
Figure 3.17. The strain rate in the TCT test was found to be about 3 s−1 in the fast
stretching zone. This strain rate is higher than the faster strain rate reached during
the uniaxial cyclic traction tests. Thus, in order to estimate the dissipated energy
in the TCT test, the Rd ratio for the 1 s−1 at each temperature will be used. This
will lead to an underestimate of the dissipated energy. At 20 ◦C, the dissipation
ratio will be Rd =85%. Obviously, this high value reflects the fact that most of the
macroscopic work of fracture is used to overcome the visco/plastic dissipation in the
stretching of the interlayer. The value of Rd is close to 90% at 10 ◦C but drops at
40% at 50 ◦C.

5.7 Influence of the temperature and applied ve-

locity on the dissipation mechanism
In order to look at the influence of the applied velocity and temperature on the near
crack work of fracture and on the bulk stretching work, the TCT test was conducted
with four different interlayer thicknesses 0.38 mm, 0.76 mm, 1.14 mm and 1.52 mm in
different conditions. At 20 ◦C, four applied velocities 1 mm s−1 , 10 mm s−1 , 20 mm s−1
and 30 mm s−1 were used (Figure 5.6(a)). Similarly at 10 mm s−1 , four different
temperatures 10 ◦C, 20 ◦C, 30 ◦C and 50 ◦C were tested (Figure 5.6(b)).

The raw data displayed in 5.6 are difficult to read. However in every case, the
increase of the macroscopic work of fracture with the interlayer thickness is linear.
Thus the slope and the extrapolated intercept at zero thickness can be extracted
from the raw data for a better visualization (Figure 5.7). The results at 30 ◦C and
10 mm s−1 are subjected, for an unknown reason, to large error bars. It is possible
that sample preparation was affected by some spurious effect. These experiment

84
should be repeated before full conclusions can be reached. Thus the slope and the
non-zero intercept extracted from these results do not give the same trends as the
others.
Macroscopic Work of Fracture Gm (kJ/m2 )

Macroscopic Work of Fracture Gm (kJ/m2 )

25 1 mm/s 10◦ C
10 mm/s 40 20◦ C
20 mm/s 30◦ C
20 30 mm/s 50◦ C
30

15
V↑
20 T↑

10
10
5
0
0
0 0.5 1 1.5 0 0.5 1 1.5
Thickness h (mm) Thickness h (mm)
(a) Effect of the velocity (b) Effect of the temperature

Figure 5.6: Effect of the velocity and the temperature on the macroscopic work of
fracture with the interlayer thickness. (a) Four temperatures 10 ◦C, 20 ◦C, 30 ◦C and
50 ◦C were tested for a given velocity of 10 mm s−1 . (b) Four velocities 1 mm s−1 ,
10 mm s−1 , 20 mm s−1 and 30 mm s−1 were tested at 20 ◦C.

In Figure 5.7(a), 2Γcrack increases from 4 kJ m−2 to 11.5 kJ m−2 when the applied
velocity increases at 20 ◦C. At low speed however, the intercept reaches a plateau
value of 4 kJ m−2 . Similarly 2Γcrack seems to increase from 1 kJ m−2 to 5 kJ m−2
when the temperature decreases from 50 ◦C to 10 ◦C at a given velocity of 10 mm s−1
(the non physical negative result found for 30 ◦C is discarded).
In Figure 5.7(b), Πbulk is increasing first with applied velocity at 20 ◦C from
3.5 MJ m−3 to 10 MJ m−3 , but decreases to 8 MJ m−3 for the higher velocity. Fi-
nally, the slope displays a global increase from 1 MJ m−3 to 23 MJ m−3 when the
temperature decreases from 50 ◦C to 10 ◦C at a given velocity 10 mm s−1 . Again,
Πbulk is higher at 30 ◦C than at 20 ◦C.

Due to the difficult reproducibility of some of these experiments, we only discuss

their global trends. We can conclude that the near crack work of fracture increases
when the velocity increases and when the temperature decreases. The same trends
are observed for the volume density of the bulk stretching work.

85
 15 30
10◦ C 10◦ C
20◦ C 25 20◦ C
30◦ C 30◦ C
10
2Γcrack (kJ/m2 )

Πbulk (MJ/m3 )
50◦ C 50◦ C
20

5 15

10
0
5

-5 0
0 10 20 30 0 10 20 30
Velocity δ̇ (mm/s) Velocity δ̇ (mm/s)
(a) Near crack work of fracture (b) Bulk stretching work density

Figure 5.7: Effect of the velocity and temperature on (a) the near crack work of
fracture 2Γcrack and (b) the bulk stretching work density Πbulk .

5.8 Discussion and Conclusion

We have shown that the macroscopic work of fracture provided during delamination
is used to stretch the interlayer up to a mean nominal strain 0TCT . This stretching
results in a stored elastic energy in the delaminated part of the interlayer but also
involves a large amount of dissipation. We have shown that the energy dissipation
occurs in two main zones. In the fast stretching zone, the interlayer is stretched
up to 60% of 0TCT over a length equal to the interlayer thickness. We have shown
that this fast stretching region can be modeled by a uniaxial tension test on the
interlayer alone. This test can be used to estimate the volume density of the bulk
stretching work Πbulk . This rough evauation gives a value (9 MJ m−3 at 20 ◦C and
1 s−1 ) which compares favorably with the experimental results (8 MJ m−2 at 20 ◦C
and 10 mm s−1 ). The remaining energy goes to the near crack region where a non
negligible amount of energy is also dissipated Γcrack . This energy corresponds to a
local deformation of the interlayer and to the rupture of the interfacial bonds. This
zone can not be reached experimentally due to its small size. However, the large
amount of dissipated energy in this zone could be related to the large strain and
strain rates that are expected to be found close to the crack tip. This hypothesis
will be studied thanks to a finite element model in the final part of this study.

The description of the problem that arises at this point enables us to understand
qualitatively the variation of the delamination force observed during the TCT test.
Indeed, the delamination can be related to a uniaxial tension test. Moreover, at
a given temperature, the strain 0TCT is constant with the applied velocity (Figure

86
4.9). Thus, the delamination force increases with applied velocity as observed in
the rheology of the interlayer (Figure 3.12). However, to predict the level of strain
reached by the interlayer the coupling between the stretch region and the near crack
region must be considered. This will question be tackled in the next chapters.
We might also be able to explain the different delamination regimes, described
in Chapter 4. At high temperatures and low velocities, the macroscopic work of
fracture is much smaller as the polymer is softer. Moreover, as the relaxation time
of the polymer are closer to the experimental time, the polymer can relax between
the remote loading point and the delamination front. This means that both the
energy needed for the front to propagate and the stress to break the interface, are
smaller. This could explain why the front buckles and that its propagation stops
leading to failure. At higher velocity, even if the delamination conditions are fulfilled,
the polymer becomes more brittle and more sensitive to initiation of defects resulting
from the sample preparation or clamp misalignment. However, in this high velocity
regime, the delamination process should not be different from the ones described in
the intermediate regime.

The estimation of the dissipated energy that can be done using the ratio of
dissipated energy from the cyclic uniaxial tension experiments is of great interest in
order to predict, the laminated glass resistance to impact. Considering this ratio is
close to 1 at 20 ◦C and for strain rates higher than 1 s−1 , the values of the macroscopic
work of fracture already give a good approximation of the dissipated energy.

Finally, the variations displayed by the near crack work of fracture and bulk
stretching work density with the temperature and the applied velocity require more
experiments to confirm the trends described previously.

87
Main results
• The macroscopic work of fracture can be divided in different compo-
nents:

– A bulk stretching work which is:

∗ For a small part, stored as elastic energy in the stretched
interlayer
∗ For the main part, dissipated in the fast stretching zone
– A near crack work of fracture which corresponds to dissipated en-
ergy due to the deformation of the interlayer in the delamination
front vicinity
– A separation work which corresponds to the energy required to
break the hydrogen bonds at the interface

• The bulk dissipation in the fast stretching zone can be modeled thanks
to a uniaxial tension test at the appropriate strain rate.

88
Chapter 6

Interface modification –
Preliminary results

6.1 Introduction

In the previous parts, a description of the delamination mechanisms was provided.

A steady state delamination regime was identified in a certain range of temperature
and applied velocity. Two different dissipation energies were measured in this regime.
In order to push further our understanding of the near crack dissipation mechanisms
and in the bulk of the interlayer during delamination, the impact of the interface on
these processes is investigated.

In order to change the work of separation, namely the molecular adhesion of the
interlayer onto glass, without altering its rheological properties, silanization of the
glass surface was used. This work has been conducted together with Raphaelle Kulis.
The silanization process will not be discussed here. It has been thoroughly described
by R. Kulis in her report [13]. The effect of silanization was first quantified by
peel experiments. Then, the weaker and stronger interfaces were tested in the TCT
experiment. We show that a change in adhesion affects the steady state delamination
regime. In the case of the stronger interface no delamination is observed whereas
for the weaker interface, it is possible to find a range of temperature and applied
velocity in which the delamination occurred in a steady state manner. This enabled
to study the effect of a decrease in the adhesion on the macroscopic work of fracture
and its different components.

89
6.2 Impact of silanization on the interface and on
the peel work of fracture
In order to get a weaker and stronger interface without changing the bulk proper-
ties of the interlayer (see section 3.3.2), two silanes were used: (3-aminopropyl) tri-
ethoxysilane (APTS) was used to enhance molecular adhesion while triethoxy(octyl)silane
(OTES) was used to reduce it (see 2.2). APTS is known to increase the molecular
adhesion between glass and PVB [46] due to its amphiphilic nature. Indeed, the
silanol group can attach on the glass surface (through a complex mix of hydrogen
bonds, electrostatic interactions and covalent bonds) and the amino group can form
bonds with the hydroxyl groups of the PVB interlayer. On the other hand, OTES
can similarly attach to the glass surface by mainly covalent bonds with the glass
silanols, but the long hydrophobic chain will reduce interactions with the PVB inter-
layer, thus decreasing adhesion [47] [48]. We will refer to these treated glass samples
as OTES treated and APTS treated samples. The laminated glass samples without
any surface treatment used previously will be referred to as non-treated samples and
used as reference.
The impact of silanization on the interface was evaluated through 90◦ peel tests.
A stiff backing is used on top of the PVB layer to prevent bulk stretching during the
tests. These tests were performed at room temperature and two different traction
velocities of 5 mm min−1 and 50 mm min−1 . During the peel test, a constant force
was measured (Figure 6.1). From this force level, the peel work of fracture was
calculated: Gc = F
w
. This work corresponds to both the rupture of the molecular
bonds at the interface and to the stretching of the PVB near the crack tip. This
peel work of fracture is plotted in Figure 6.2. From these two figures it appears
clearly that the APTS and OTES treatments result in the expected changes of the
interfacial adhesion.
At 5 mm min−1 , APTS treatment leads to an increase in the mean peel force from
20±4 N to 71±8 N which corresponds, for a sample width of 2 cm, to an increase in
the peel work of fracture from 0.97±0.2 kJ m−2 to 3.5±0.4 kJ m−2 . At 50 mm min−1 ,
the force increases from 41±5 N to 97±4 N which corresponds to an increase in the
peel work of fracture of 2.1±0.25 kJ m−2 and 4.9±0.2 kJ m−2 .
On the other way, OTES treatment leads to a decrease in peel force. At 5 mm min−1
the peel force decreases from 20±4 N to 3±1 N. At 50 mm min−1 the peel force de-
creases from 20±4 N to 5±2 N. This corresponds to lower values of the energy release
rate of 0.17±0.05 kJ m−2 and 0.26±0.1 kJ m−2 respectively.

90
 80

60
Peel Force (N) APTS
40 Standard
OTES

20

0
0 20 40 60 80
Time (s)

Figure 6.1: Peel force measured at 5 mm min−1 , at room temperature on non-treated

PVB/Glass sample (without silane modification of the interface), on APTS treated
sample (enhanced adhesion) and OTES treated sample (reduced adhesion).

7
Peel Work of Fracture (kJ/m2 )

APTS
6 Standard
OTES
5

0
0 20 40 60
Peel speed (mm/min)

Figure 6.2: Peel work of fracture found in the peel test at two velocities 5 mm min−1
and 50 mm min−1 for the non-treated, OTES treated and APTS treated samples.

91
6.3 Impact of an interface modification on the
TCT test response

6.3.1 Different steady state delamination regimes

Impact of the two silane treatments on adhesion being established, the modified
surface were assembled with PVB interlayer and tested in TCT geometry. The
samples with modified adhesion were first tested in the steady state regime found for
the non-treated surface at 20 ◦C and 10 mm s−1 . We observed qualitatively different
behaviors.
On APTS treated samples, no delamination was observed and the interlayer
broke immediately (Figure 6.3(a)). On OTES treated samples, delamination was
observed initially but the delamination front undulated and presented an unstable
behavior. The delamination stopped and the polymer ruptured (Figure 6.3(c)).
This behavior was similar to the one observed at high temperature or lower imposed
velocities on the non-treated TCT sample.
For the OTES treated samples, the steady state delamination regime was recov-
ered at higher velocities or at lower temperatures. It has been found at 20 ◦C for
velocities higher than 30 mm s−1 that the delamination was stable. At 10 ◦C all the
velocities tested in between 10 mm s−1 and 100 mm s−1 displayed stable delamina-
tion.
On the contrary, for the APTS treated samples, no delamination was observed
for any of the temperature and the velocities applied.
Thus it appears that a weaker interface shifts the steady state delamination
regime towards the lower temperature and higher velocities. Here, the increase in
adhesion was too large to observe any delamination. A possible hypothesis is that
the non-treated sample is already close to the upper limit of adhesion and that a
small change of the interface results in the rupture of the interlayer prior to any
delamination. This hypothesis could be tested by using different concentrations of
the APTS solution or by using silane mixtures.

6.3.2 Steady state delamination for the lower adhesion

For the weaker interface the steady state delamination regime was tested for different
velocities and temperatures.
Some general observations can be made first. The delamination fronts on the
back glass and on the front glass were often not propagating at the same velocity.

92
 (a) APTS

(b) Not treated

(c) OTES

Figure 6.3: The different behaviors observed during the TCT experiment at 20 ◦C
and 10 mm s−1 . The APTS displays an immediate brittle rupture with no delam-
ination, while the OTES shows delamination with unstable and undulating fronts
before rupture.

93
Thus the two delamination fronts are visible as displayed in Figure 6.4. Further-
more, in some cases, an abrupt delamination occurred on one glass pane followed
by a stable and steady state delamination on the other glass. At 10 ◦C and for an
applied velocity of 20 mm s−1 , the abrupt delamination occurs at a velocity close
to 200 mm s−1 and the slower delamination at a velocity about 30 mm s−1 . In the
case where an abrupt delamination is observed, only the steady state delamination
front was recorded and the measured strains and delamination forces were similar
to steady state delaminations. The experiments conducted at 10 ◦C on the OTES
samples are more reproducible than those conducted at 20 ◦C.

Fast delamination fronts

on the back glass

(a) Propagation after 25 ms

Fast delamination front

on the back glass

Slow delamination front

on the front glass

(b) Propagation after 850 ms

Figure 6.4: TCT test on an OTES treated sample at 20 mm s−1 and 10 ◦C. Sample
width is 50 mm (a) Abrupt delamination on the back glass at a velocity close to
200 mm s−1 . (b) The second front propagates more "slowly" at a velocity about
20 mm s−1 .

In Figure 6.5, the delamination velocity, force and strain found for the different
temperatures and velocities for the OTES samples are compared to the non-treated
reference. First of all, the delamination occurs at a much faster velocity for the

94
OTES samples than for the non treated ones. The strain is found to be much
lower for the OTES treated sample than for the non treated ones at a 20 ◦C for a
given velocity. The force is similar at 20 ◦C for both the OTES and the non treated
samples at 30 mm s−1 and 60 mm s−1 .
Similarly to what has been found for the non-treated samples (see Section 4.3),
the force increases with the applied velocity and decreases with the temperature
for the OTES samples. At 20 ◦C the force increases from 450 N to 600 N between
30 mm s−1 and 60 mm s−1 . The experiment at 100 mm s−1 can be questioned due
to the fast breaking of the interlayer after a small stable delamination. At 10 ◦C
the increase in force is clearer from 547 N to 847 N when the velocity increases from
10 mm s−1 to 100 mm s−1 . The strain does not display the same behavior as for the
non treated sample. Indeed, if the strain increases with the velocity from 0.9 to 1.4,
these values are not changing with the temperature unlike what we have observed for
the non-treated sample strain (see Section 4.3). The delamination velocity displays
the same trends. It increases from 5 mm s−1 to 32 mm s−1 at 20 ◦C as the applied
velocity increases from 10 mm s−1 to 100 mm s−1 but it does not change much as the
temperature decreases at 10 ◦C.
Due to the decrease of the strain, the macroscopic work of fracture is lower in
the case of the weak adhesion at 20 ◦C. This effect is more visible as the velocity
increases (Figure 6.6).

6.3.3 A change in the dissipated energies

The OTES treated glass can also be tested in the TCT test with different interlayer
thicknesses. In Figure 6.7, three different thicknesses were tested in a reproducible
manner (0.38 mm, 0.76 mm and 1.14 mm) at 10 ◦C and 30 mm s−1 . These conditions
were chosen because they lie in the steady state delamination regime. However, the
test is still difficult to perform as at this temperature and velocity the interlayer is
sensitive to any experimental defect. Thus, only one point has been retrieved for
the 1.52 mm thickness and one could argue that this result is not reproducible. If
the first three points only are considered, the macroscopic work of fracture is found
to be linear with the interlayer thickness as it was found for non treated sample. At
this temperature and for this applied velocity, 2Γcrack is about 9 kJ m−2 and Πbulk
about 5.2 MJ m−3 .

95
 40
OTES 10◦ C
Delamination Speed (mm/s)

OTES 20◦ C 800

Non treated 20◦ C
30

(N)
600

Force FTCT
0
20
400

10 200 OTES 10◦ C

OTES 20◦ C
Non treated 20◦ C
0 0
0 25 50 75 100 100 101 102
Velocity (mm/s) Velocity δ̇(mm/s)
(a) Delamination velocity (b) Delamination force
2.5

2
Strain 0TCT

1.5

0.5 Non treated 20◦ C

OTES 10◦ C
OTES 20◦ C
0
100 101 102
Velocity δ̇(mm/s)
(c) Strain

Figure 6.5: The delamination velocity, force and strain measured during the TCT
tests performed on the OTES treated samples at 10 ◦C and 20 ◦C compared, for
different velocities, to the non-treated samples results found at 20 ◦C.
.

96
 Macroscopic Work of Fracture Gm (kJ/m2 )
35
Non treated 20◦ C
OTES 10◦ C
30
OTES 20◦ C
Non treaded steady state
25 OTES treated steady state

20

15

10

0
100 101 102
Velocity δ̇(mm/s)

Figure 6.6: The macroscopic work of fracture during the TCT tests performed on
the OTES treated samples at 10 ◦C and 20 ◦C compared, for different velocities,
to the non-treated samples results found at 20 ◦C. The steady state delamination
regime is shifted towards higher velocities and lower temperatures as the interface
is weaker.
Macroscopic Work of Fracture Gm (kJ/m2 )

20

15 ?

5.2 MJ/m3
10

5
8.9 kJ/m2

0
0 0.5 1 1.5
Interlayer Thickness h (mm)

Figure 6.7: Dependence of the macroscopic work of fracture on the interlayer thick-
ness during a TCT test performed on an OTES treated laminated glass at 10 ◦C and
30 mm s−1 .

97
6.4 Discussion
A change of the interface leads to a shift of the steady state delamination conditions
as displayed in Figure 6.6. More specifically it was found that a weaker interface
leads to a shift of the steady state delamination regime towards lower temperatures
and higher velocities. One could assume that an increase of toughness will shift the
steady state delamination regime towards higher temperatures and lower imposed
velocities. However, this assumption has still to be verified. The increase of adhesion
provided by the APTS treatment is too large to allow any delamination. One could
reduce the APTS concentration to limit the increase in adhesion.
The decrease of the glass/interlayer adhesion due to the OTES treatment was
compared to the previous results found on the non-treated samples in terms of force
and strain during the delamination. It leads to the conclusion that a weaker inter-
face is responsible for a lower macroscopic work of fracture due to the smaller strain
necessary to delaminate the polymer. The variation of the macroscopic work of frac-
ture with the interlayer thickness enables us to push forward this comparison. The
result found at 10 ◦C and 30 mm s−1 on the OTES treated sample can be compared
with the ones found at 10 ◦C and 10 mm s−1 on the non treated sample (see Section
5.7).
The volume density of the bulk stretching work is lower for the OTES treated
sample (5 MJ m−3 ) than the one found on the non-treated sample (22 MJ m−3 ) (Fig-
ure 6.7). Moreover, the OTES treated sample was tested at a higher imposed velocity
of 30 mm s−1 and Πbulk tends to increase with the applied velocity. Thus we might
conclude that a weaker adhesion induces a decrease of the bulk stretching work.
This can be related to the lower mean nominal strain found previously. The inter-
layer being less stretched, the dissipation associated to this stretch is lower even if
the strain rate could be more important.
The near crack work of fracture, Γcrack = 9 kJ m−2 , was found to be higher for
the OTES treated sample than the one one found on the non-treated sample at a
lower velocity (5 kJ m−2 ). As the near crack work of fracture tends to increase with
applied velocity, it is difficult to tell if there is a significant difference with the non
treated sample. However, it is interesting to see that despite the weaker interface,
Γcrack is not highly reduced. This can be explained by the faster front propagation
observed on the OTES treated samples (Figure 6.5(a)), which might lead to higher
strain rates that can compensate the lower stain level reached by the interlayer in
this zone.
Another set of experiments at 10 ◦C and 30 mm s−1 has to be conducted on the

98
non-treated samples with different interlayer thicknesses to quantify the bulk and
close to the front dissipation and to confirm these results. It is however interesting to
notice that results similar to the one displayed by the OTES treated sample seems to
arise from TCT experiments conducted on samples with a nano meter scale metallic
oxyde multilayer in-between the glass and the PVB. This metallic oxide multilayer,
is more brittle than the non-treated interface and the rupture occurs inside the
metallic layer. This other preliminary results suggest that the observations made
during this study can be extended to other systems.

6.5 Conclusion
These preliminary results have shown that interface modification affects the steady
state delamination of the interlayer. Especially, for a weaker interface, a faster front
velocity and a lower mean nominal strain have been recorded. A steady state de-
lamination regime has been identified at higher velocities and lower temperatures
compare to the non treated reference. This lower adhesion level also affects the bulk
stretching work and the near crack work of fracture. However more experiments
are required to compare these results with the non treated reference samples. For
example, the effect of the interlayer thickness on the macroscopic work of fracture
has to be checked. APTS treatment has induced an increase in adhesion for which
no steady state delamination has been observed. Using lower concentrated APTS
solution or silanes mixtures could be a way to mitigate the increase in adhesion. Fi-
nally, the effect of interface modification on laminated glass response during impact
should be investigated.

99
Main results
• Silane treatments resulted in the expected changes of adhesion: the
OTES treatment reduces it while the APTS treatment increases it.

• A change at the interface affects the range of the stable delamination

regime:

– When the adhesion increase was too large, no steady state de-
lamination was observed for any of the tested velocities and tem-
peratures. A more moderate increase should be considered.
– The decrease in adhesion shifts the steady state regime towards
higher velocities and lower temperatures.

• Compared to the non-treated reference sample, the lower adhesion

level displays:

– A lower constant mean nominal strain TCT .

– A higher delamination front velocity.

• As a result the decrease in the macroscopic work of fracture is mainly

to be found in the bulk dissipation:

100
Chapter 7

Finite element modeling

description

7.1 Introduction

In the previous chapters we have described the delamination process of the interlayer
and the associated dissipation of energy (see chapters 4 and 5). Especially, we
have shown that the stretching of the interlayer dissipates a large part of the total
dissipated energy in two zones: a fast stretching zone and a near crack area (5.3).
We have experimentally been able to study the fast stretching zone but the vicinity
of the delamination front is, mainly due to its size, out of reach. Moreover, it is also
difficult by the experiment to understand how these different zones are connected
together and to the remote loading. Thus a finite element model of the TCT test has
been developed to answer these different questions. It can also be a useful tool to
confirm and study the effect of several parameters such as applied velocity, adhesion,
interlayer behavior and sample geometry on the delamination response.

In this chapter, after a brief introduction of the cohesive zone model used to
represent the interfacial rupture between the interlayer and the glass, a complete
description of the model is given. This finite element model is then used to reproduce
qualitatively the experimental results. The existence of the two dissipation zones
will be confirmed and the processes near the crack will be more closely studied.
Finally, the effect of viscoelastic relaxation times and adhesion parameters will be
observed.

101
 PVB elements

Cohesive elements

Figure 7.1: Cohesive zone layer of elements representing the interface. The size of
the cohesive elements (10−3 ) is much smaller than the size of the PVB layer above
(1).

7.2 Cohesive zone model for the interfacial rup-

ture
Cohesive zone models have been used in several papers dealing with fracture in bulk
polymers or at the interface between a soft polymer and a rigid substrate [49] and
more specifically between glass and PVB ([50] or [12]). These cohesive zone models
are used to describe the work of separation of the interface at the molecular scale.
One model to describe the adhesion at an interface is the Dugdale and Barenblatt
cohesive zone model [51]. It is a simple model to describe adhesion. The stress σ is
supposed to be constant equal to σ0 in the cohesive zone while the distance between
the two surfaces ∆ remains smaller than a critical separation distance ∆s :

σ = σ0 if ∆ < ∆s (7.2.1)

When ∆ reaches ∆s the interface fails and σ = 0:

σ = 0 if ∆ ≥ ∆s (7.2.2)

More complex models taking account mode mixity (while the original Dugdale
model is only for mode I opening) were developed such as in [52] or [53]. The model
implemented in ABAQUS was developed by [54]. It implements irreversible damage
of a layer of elements representing the interface (Figure 7.1). Moreover it takes
into account mixed mode loading. The following explanation is based on the one
presented by Barthel and coworkers [55].
In this model, the cohesive elements sustain a certain amount of deformation with
an elastic stiffness K as shown in equation 7.2.3 where the stress T and displace-
ment ∆ can be decomposed in a normal component (T22 and ∆22 ) and a tangential
component (T11 and ∆11 ).

T = K∆ (7.2.3)

102
 The combined action of normal (direction 2) and tangential (direction 1) loading
in mixed mode loadings can lead to the initiation of rupture below the uniaxial
tension level. To take this effect into account, a quadratic rupture initiation criterion
is defined. T11
0
and T22
0
are parameters of the cohesive zone that have to be chosen.
The peak stress T 0 is then the solution of:
!2 !2
T22 T11
0
+ 0
=1 (7.2.4)
T22 T11
In the case where equation 7.2.4 is verified, the mode mixity angle is defined by:

T11
 
tan(ψ) = (7.2.5)
T22
This enables the definition of the work of separation as a function of ψ (equation
7.2.6 with η defined by ΓII
s = Γs (1 + tan (ηπ/2))).
I 2

Γs (ψ) = ΓIs (1 + tan2 (ηψ)) (7.2.6)

This elastic behavior holds up to the peak stress T 0 (corresponding to ∆0 ). After

this peak is reached, damage takes place which decreases the stiffness of the cohesive
elements (equation 7.2.7).

T = K(1 − d)∆ (7.2.7)

d is the damage parameter which is defined by equation 7.2.8. ∆ = ∆21 + ∆22 is

q

the scalar displacement, ∆0 is the scalar displacement at the maximum stress and
∆s is the scalar critical displacement.

∆s ∆ − ∆ 0
d= (7.2.8)
∆ ∆s − ∆0
Above the critical displacement ∆s the crack opens. This critical displacement
is defined by equation 7.2.9.

1
Γs (ψ) = ∆s T 0 (7.2.9)
2
In other words, different parameters have to be defined in ABAQUS:

• the stiffnesses K11 and K22 .

0
• T11 and T22
0
which define the traction separation criterion

103
 • the work of separation Γs or the critical opening displacement ∆s . This value
can be defined for different mode mixity angles ψ.

7.3 Model description

In this part we are going to describe the conditions and parameters applied on the
TCT geometry. Only one quarter of the geometry is modeled due to symmetries
of the problem (Figure 7.2). In order to keep the notation understandable and
compatible with the ABAQUS notation, the horizontal direction will be referred to
as direction 1 and the vertical direction will be referred to as direction 2.

Model zone

Figure 7.2: Due to symmetries of the problem only a quarter of the TCT sample is
modeled.

The model units are taken in order to get proper dimensionless numbers. The
distance are all in mm. Hence all moduli and stresses are in MPa. Finally all times
are in s.

Geometry

The model geometry is presented in Figure 7.3. First the length of the sample has
to be sufficiently long so a steady state delamination regime can be observed. This
is obtained for a length larger than the sample thickness. As we choose here h/2 = 1
this stable delamination regime will be obtained for a length larger than 2. Moreover,
in order to avoid interference between the crack tip field and the left boundary the
sample is 15 long. The cohesive zone height has no theoretical meaning as cohesive
elements are 1 dimensional elements. However, it has been found that cohesive
elements should not deform too much for the calculation to converge without issues.
Thus a thickness of 10−2 was used. Finally, as the cohesive elements are not designed
to model crack initiation a small initial crack was inserted in the model. This crack
length is 1. This length will be discussed in the next part.

104
 In order to avoid contact definition problems in-between the cohesive zone and
the PVB interlayer, a single part was drawn. This part was then cut into two
“materials”.

1
15

1 PVB
Cohesive Zone
10−2

Figure 7.3: Dimensions of the finite elements geometry.

Boundary conditions

The loading and boundary conditions applied on the previous geometry are displayed
in Figure 7.4. The displacement on the left part of the interlayer are blocked so
that no delamination is observed on this part of the interlayer (except when the
delamination fronts reach it). The top part of the geometry corresponds to the
middle plane of the TCT test sample. Thus, a plannar symmetry along direction
1 is applied on this border. It corresponds to no displacement in direction 2 and
no rotations. On the bottom of the cohesive zone, all displacements and rotations
are banned. We assume indeed that the glass is rigid and does not move nor bend
during the experiment. This is also the reason why we do not model the glass.
Finally, a displacement is applied on the right part of the interlayer. The maximal
displacement and the overall loading time are defined. In the viscoelastic models
this time will define the applied velocity.

Plannar symmetry along direction 1

No displacement
δ and δ̇
in direction 1

Fixed to glass

Figure 7.4: Boundary and loading conditions

The initial crack length is in close relationship with this boundary conditions. As
we are pulling on the right hand side of the interlayer, if the initial crack length was
null, the initial pulling will be pure mode II. In this case, the interlayer elements,
close to the cohesive zone, deform too much and convergence fails. This is why

105
the initial crack length should be long enough so the mode mixity of the initial
pulling does not lead to a critical deformation of the these elements. However, as
this free interlayer length will be deformed prior to any extension of the cohesive
zone elements, the first steps of the calculation consist in some sense in an uniaxial
tension test. The initial crack length will define the strain rate applied on this
interlayer free part. Hence, because of the viscoelastic response of the interlayer, if
this crack length is too large, the strain rate will be too small to reach the critical
stress for the cohesive zone to break, which in turn will lead to a severe overshoot
in the steady state strain.
In conclusion, the initial crack length has to be short enough to reach the critical
stress for the cohesive elements to break at a reasonable strain and long enough so
the mode mixity is far from pure mode II. We have found that a length equal to
half the interlayer thickness is adequate.

Meshing

The mesh was chosen so the mesh size is small close to the cohesive zone and large
far from it. Each edge of the model is seeded with the proper mesh size and the
ABAQUS meshing software was used to compose the mesh on the whole part.
The cohesive zone mesh is made from a single layer of quadratic elements COH2D4.
These elements have a width of 0.007 and a thickness equal to the cohesive zone
thickness 10−2 . There is one cohesive element below each PVB element.
The PVB elements are quadratic plane strain elements CPE4RH. The PVB layer
is divided in two parts. A first band, just above the cohesive zone, is finely meshed
with elements of width 0.007. This layer has a height equal to 1/25 and there is
six PVB elements layers in this fine mesh band. Above this first zone the meshing
gets progressively looser as it goes away from the interface. A gradient of mesh is
imposed vertically from 0.007 to 0.01. On the top edge of the interlayer a seed of
size 0.1 is imposed.
The different components of this mesh are presented in Figure 7.5.
The mesh dependence was tested by doing the same calculation with different
mesh sizes. In Figure 7.6, the results for the crack tip position and the strain applied
far from this crack tip are given for four calculation with different mesh sizes: the
standard mesh corresponds to the one described above, the sizes of all elements were
multiplied by 2 and 4 respectively for larger mesh sizes and divided by two for a
finer mesh size. When the mesh size increases, a significant difference in the result
can be seen. The larger the mesh is, the larger the difference is. When the mesh

106
 Large mesh 0.1

Mesh gradient from

0.007 to 0.1
6 elements layer
in fine mesh zone

Fine mesh 0.007 in PVB and cohesive zone

Figure 7.5: Standard mesh used in the TCT test finite elements simulation

size is twice smaller, no difference was seen in the result, but the calculation time
increased greatly. Thus, it was found, that the standard mesh size presented just
before was the larger which enabled good precision with minimal calculation time.

8
1.5
Crack Tip Position (mm)

6
1
Strain 11

x4 0.5
2 x4
x2 x2
Std Std
÷2 ÷2
0 0
0 0.5 1 0 0.5 1
Time (s) Time (s)
(a) The crack tip position (b) The strain applied far from the crack tip

Figure 7.6: Variation of the results of the FEM calculation depending on mesh size

Convergence problems: artificial dissipation

Even if the interlayer is dissipating energy through viscoelasticity it was found neces-
sary to use an artificial dissipation inside the cohesive zone elements. This artificial
viscosity η helps the convergence of the calculation by adding a relaxation time for
these elements (see Abaqus Documentation Viscous regularization in Abaqus/Stan-
dard Section 32.5.6). As explained by Gao [56], this viscosity will regularize the

107
stress path followed during the damage of a cohesive element. Thus the energy dis-
sipation in the cohesive zone is no longer the static energy release rate Γs but an
effective adhesion energy Γ∗s which will have to be evaluated.

Another convergence problem has often arisen when the strain gradient is too
large on some elements. This problem especially arises during the first part of the
experiment were the initial crack part of the interlayer is loaded up to the point
were the crack initiates. To solve this problem two actions have been taken.
First, the tolerance value on the strain has been defined high enough for the
convergence to occur. The CETOL parameter has thus been defined at 10−2 . This
error on strain was defined as suggested in the ABAQUS documentation by consid-
ering a admissible stress error and a typical elastic modulus. The modulus of the
material is about 10 MPa. This gives an error on stress about 10−1 MPa. Stresses
during the calculation are of the order of 10 MPa. The error was thus considered to
be acceptable. Moreover, the documentation states that in the case of a viscoelastic
mode a higher value for the CETOL parameter can be chosen.
Secondly, in order to facilitate the initiation steps, an adhesion gradient was
defined at the beginning of the cohesive zone. On a distance of 0.1 mm, the adhesion
increases up to the chosen value of adhesion. All the cohesive zone parameters
including the initial stiffness K 0 and the maximal stress T 0 are changing along the
gradient which is defined with a hyperbolic tangent function: f (x) = tan(x/0.1).
On the rest of the cohesive zone part the adhesion is constant. The gradient distance
is chosen so the change in adhesion is steep enough not to affect the results and long
enough to ease the initiation convergence issues.

Materials

Now that we have defined the geometry and the mesh plus several other parameters
concerning the elements, we will define the material parameters used in this system.

Typical parameters of the cohesive zone are chosen with no mode mixity
difference thus these quantities are defined similarly in direction 1 and 2. First
a maximal deformation of the cohesive element before rupture was chosen equal
to 100%, which means that the maximal displacement of the cohesive element is
∆s = 10−2 (mm). Then we choose the work of separation equal to 38 J m−2 . In
our sets of units Γs is thus 0.038 (MPa mm). This value has been ajusted so the
stretching of the interlayer and the delamination velocity are comparable to the
experimental values for an applied displacement speed of 10 mm s−1 and at 20 ◦C.

108
The total energy dissipated by the cohesive element deformation before breakage
is however affected by the viscosity introduced in the cohesive zone elements. This
viscosity is set to η =0.001 s−1 . The effective energy required to stretch and separate
the cohesive elements Γ∗s , will be determined from the calculation results.
The critical stress is set to T 0 = T11
0
= T22
0
equal to 7.5 (MPa). Note that the
normal direction is direction 2 while the tangent direction is direction 1. In order
to get a stiff element without convergence problem it is commonly admitted that
damage should begin for a displacement value ∆0 such that ∆0 = ∆s /100 = 10−4 .
This leads to K 0 = K11 = K22 = 7.5.104 (MPa mm−1 ). The effective values of T 0
and K 0 , respectively T ∗ and K ∗ , are also determined for every calculation.

Concerning the interlayer, we have seen how complex its behavior is in Chap-
ter 3. Here we will use a simple model which will keep some of the main features
of the interlayer behavior. The time dependence of the material behavior will be
described using linear viscoelasticity combined with a non-linear hyperelastic model:
a generalized Maxwell model of the viscoelasticity and an Arruda-Boyce model of
the hyperelasticity. We will only use one relaxation time for the viscoelastic part.
To use such a simple model will, in a first time, help us assess the finite element
modeling while qualitatively retrieving some of the experimental observations.
The viscoelastic model is defined through the shear modulus relaxation function:

N
" #
µ(t) = µ0 1 + αi (exp(−t/τi ) − 1) (7.3.1)
X

i=1

where µ0 is the instantaneous modulus, τi are the relaxation relaxation times

and αi the associated coefficients. Here we use a single relaxation time (Zener
Model) τ1 = 0.01 s and α1 =0.9. The instantaneous modulus is µ0 =10 MPa. Using
these material parameters, the storage and loss modulus as well as tan δ ratio are
plotted in Figure 7.7. This shows that there is a ratio of 10 in between the relaxed
and instantaneous modulus and that the dissipation is maximal around 0.01 s−1 .
The relaxation of the shear modulus, is also plotted in Figure 7.8. This model
qualitatively represents the behavior observed in DMA experiments (Figure 3.6).
The hyperelastic Arruda Boyce model is defined by:

1 5
iCi I1 (t)i−1
!
σ(t) = 2µ(t) λ(t) − (7.3.2)
X
λ(t)2 i=1 λ2(i−1)
m

In this equation, the shear modulus is replaced by the viscoelastic model defined
previously in equation 7.3.1 and the maximal chain extensibility will be set at 100%

109
 102 1.5
Storage Modulus
Loss Modulus
101 tan(δ)

100 1
Moduli (MPa)

tan(δ)
10−1

10−2 0.5

10−3

10−4 0
10−3 10−2 10−1 100 101 102 103 104 105
Frequency (Hz)

Figure 7.7: The storage and loss moduli as functions of frequency. The glass tran-
sition is observed around 100 s−1 and corresponds to the maximal dissipation.
Shear modulus µ(t) (MPa)

10 µ0 = 10MPa

4
τ1 = 0.01s µ∞ = 1MPa
2

0
0 0.05 0.1 0.15 0.2 0.25
Time (s)

Figure 7.8: Relaxation function of the shear modulus for τ1 =0.01 s and α1 =0.9.

110
of strain (λm =2).
These different parameters were chosen to dissipate an large amount of energy at
certain strain rates between 10 s−1 and 100 s−1 . These strain rates will be activated
close to the crack tip and in the bulk deformation zone. In Figure 7.9 one can see that
at these strain rates, the dissipation is quite large and the material displays a strong
strain hardening at strain larger than 100%. However, as soon as the strain rates
are slower or higher, this model of the material tends towards a pure hyperelastic
model in which there is little to no dissipation. Especially, at high strain rates, we
have seen that this does not reproduce the real PVB behavior (Figure 3.12). The
next step in the modeling work will have to take into account the viscoplastic part
of the material description.

200
Strain rate
1000 s−1
150 100 s−1
True Stress (MPa)

10 s−1
100 1 s−1

50

-50
0 0.5 1 1.5 2
Nominal Strain
Figure 7.9: Calculated traction curve for our model polymer with µ = 10 MPa,
λm = 2 and one relaxation time τ1 = 0.01 s and α1 =0.9.The calculation was made
for four different strain rates from 1 s−1 to 1000 s−1 . These results can be compared
with experimental curves in Figure 3.12.

7.4 Recovering a steady state delamination

The model extensively described in the previous part will now be used to recover
qualitatively some experimental results. Especially a steady state delamination
will be obtained and the main characteristics of the delamination and interlayer
stretching will be extracted in this regime.

111
7.4.1 Decohesion processes
First, to check that the cohesive element response gives consistent results, we com-
pare it to the expected theoretical response describe in 7.2. In Figure 7.10, different
curves are plotted as a function of the distance to the delamination front. The
stresses (measured at the top of the cohesive zone) in the two directions far from
the front are equal to the hydrostatic value. When they reach the value T 0 one can
notice that the damage coefficient d (in the cohesive elements) starts to increase.
However, the stresses keep increasing after reaching T 0 up to T ∗ due to the viscosity
introduced in the cohesive elements. When the damage coefficient reaches 0.99 the
crack opening (position of the top nodes of the cohesive zone) is equal to the critical
value ∆s = 0.01 mm and the cohesive elements break. Instead of going straight to
zero, the stress relaxes due to the viscoelastic nature of the interlayer.
20 0.02 1
d
∆
σ11
σ22
15 0.015 0.75

Damage parameter d
Opening ∆ (mm)
Stresses (MPa)

∆f =0.01mm

10 0.01 0.5
T0 =7.5MPa

7.5

5 0.005 0.25

0 0 0
-0.2 -0.1 0 0.1
Distance to front (mm)

Figure 7.10: The stresses in the two directions and the position of the top nodes of
the cohesive zone are plotted as functions of the distance to the delamination front.
The damage parameter measured in the cohesive elements is also plotted.

In Figure 7.11, the three components of the stress tensor, measured at the top
nodes of the cohesive zone, are plotted as functions of the crack opening. The shear
stress is close to zero. This representation also shows that the cohesive element
response is linear far from the delamination front where displacement of the cohesive
elements node is below ∆0 = 10−4 mm. The stiffness of the elements in this region
can be measured and is equal to K ∗ = 7.5.104 MPa mm−1 as prescribed. For an

112
opening greater than ∆0 , the cohesive elements are damaged as shown just before.

10
σ22
σ11
8 σ12

Stresses (MPA)
6

K=7.5e+04 MPa/mm
4

0 0.05 0.1 0.15 0.2 0.25

Opening ∆ (µm)

Figure 7.11: The different stress components are plotted as a function of the crack
opening. The stresses σ11 and σ22 evolve linearly with the crack opening up to the
opening ∆0 =0.1 µm. A stiffness K ∗ = 7.5e4MPa mm−1 is measured.

7.4.2 Hydrostatic stress induced by the boundary condi-

tions and the incompressibility

One can notice in the previous picture that the stresses in directions 1 and 2 do not
decrease to zero in the region which is not delaminated. This is due to the facts
that the interlayer is described as an incompressible material and that boundary
conditions imply that no displacement are permitted far from the delamination
front. Thus hydrostatic stresses appear in the material. In Figure 7.12, the stress
in direction 1 (pulling direction) in the material is plotted along the median plane
as a function of the distance to the delamination front. One can see that the stress
value is constant before the front and equal to 4.3. This value is the same for
the stress in direction 2 (orthogonal to the pulling) before the front. The strains
associated with this hydrostatic stress are of the order of 11 = 10−4 , 22 = 10−5 and
12 = 10−4 . Thus energy associated to this hydrostatic component can be evaluated
and is about Gh ≈ 10−4 kJ m−2 . Even if the hydrostatic stress is not negligible, the
energetic contribution of this stress can be neglected compared to the other energies.

113
 10

Bulk Stresses (MPa)

8

6 σ11
σ22
4 σ12

0
-5 0 5 10 15 20
Distance to front (mm)

Figure 7.12: The different stress components along the median axis. Before the
delamination front, a significant hydrostatic stress is observed.

7.4.3 Energy flows balance

In order to ensure that the calculation gives consistent results, one can look at the
different energies recorded during the delamination. These energies are plotted as
functions of the delamination front position (Figure 7.13). These plots show that
a steady state delamination is reached as the energies are linear functions of front
position. The energy release rates associated to these energies are calculated by
taking the slope of these curves divided by the interlayer width equal to 1 mm. The
different energy release rates are displayed in Table 7.1.
The sum of all energies is equal to the macroscopic work of fracture. This reflects
the fact that the system reached a steady state in a consistent manner.
Secondly, the dissipated energy in the cohesive zone, corresponding to the viscous
and damage dissipation in the cohesive elements, is dominated by viscous dissipation.
This dissipation is much larger than the value of Γs that was defined in the model.
We can thus define an effective adhesion energy Γ∗s equal to 0.2 kJ m−2 . Because the
interlayer does not dissipates as much energy as the real material (µ∞ = µ0 /10),
the viscous dissipation and the elastically stored energy in the interlayer are of the
same order of magnitude. The bulk stretching work Γbulk is equal to the sum of this
elastic stored energy and viscoelastic dissipated energy in the interlayer: Γbulk =
Γvisco
bulk + Γbulk . The viscoelastic dissipation is sufficient to enhance the adhesion.
el

Indeed, the energy required to break the cohesive elements is equal to 0.2 kJ m−2 and
a macroscopic work of fracture equal to 5.9 kJ m−2 is required for the delamination

114
 60
Macroscopic Work of Fracture
50 Interlayer Viscoelastic Losses
Interlayer Elastic Energy
CZ Viscous Dissipation
Energy (mJ) 40 CZ Damage Dissipation

30

20

10

0
0 2 4 6 8
Front Position (mm)

Figure 7.13: Different energies recorded during the delamination. A steady state
delamination is reached as the energies increase linearly with the delamination front
position. The macroscopic work of fracture Gm (blue) is the total work provided to
the system. Both viscoelastic losses (red) and elastically stored (yellow) energies in
the interlayer are plotted. The viscous (purple) and damage (green) dissipation in
the cohesive zone (CZ) are also plotted.

Energies kJ m−2

Macroscopic work of fracture 5.9

Elastic energy stored in the interlayer 2.5

Viscoelastic dissipation in the interlayer 3.2

Viscous dissipation in the cohesive elements 0.2

Damage dissipation in the cohesive elements 0.0026

Table 7.1: The different energies measured during the steady state delamination.

115
front to propagate. This macroscopic work of fracture corresponds to the energy
brought to a quarter of the TCT sample and is similar to the values found in the
experiments for half a sample (≈ 10 kJ m−2 – see Figure 5.1).

7.4.4 Far field measurements

Several results will now be qualitatively compared to the experiment. The point is
to ensure that the results found here are representative of what is occurring during
the experiment. Even if the material behavior is only partially described and even if
the effective adhesion energy is much larger than the energy provided by hydrogen
bond adhesion between PVB and glass, the far from the delamination front behavior
can be recovered.
First, the delamination front propagates at a constant velocity about 7 mm s−1 .
This value is of the same order of magnitude as the one found during the experiment
(see Figure 4.5).
After a first transient state, the delamination occurs in a steady state manner.
The strain on the right end side of the PVB is equivalent to the strain measured
in between the delamination fronts during the TCT Test. This strain in direction
1 (pulling direction) is plotted, in Figure 7.14(b), as a function of time. One can
see that the strain measured in direction 1 will reach a plateau value of 1.4 after
approximately a time of 0.4 s. This value, corresponds to a delamination over a
length of approximately 2 mm equal to the interlayer thickness (Figure 7.14(a)).
The strain value is close to the one found in the TCT experiment (see Figure 4.6).
The initial transient behavior spans over a length equal to the interlayer thickness
as in the experiment (see Figure 5.2).
The relation between the applied displacment δ, the strain  and the delamination
front position a is here given by  = δ/a as we pull on half a TCT sample.
The stress in direction 1 on the right end side of the interlayer is plotted as a
function of time in Figure 7.15. This stress reaches a steady state value of 10 MPa
which lead to a force of 1000 N for an interlayer thickness of 2 mm and 50 mm width.
As the interlayer thickness is only a parameter of the calculation, the force for an
interlayer thickness of 0.76 mm can also be deduced from this stress value and would
be about 380 N. These values are of the same order of magnitude as the one found
in the experiment (see Figure 4.4).

116
 Delamination Front Position (mm)
8

0
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)
(a) Delamination front position.

1.5

1
Strain 11

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)
(b) Strain in direction 1 on the right end side of the inter-
layer.

Figure 7.14: (a) The delamination front position is plotted as a function of time.
Delamination velocity is about 7 mm s−1 . Compariason with experiment. (b) The
strain in direction 1 on the right end side of the interlayer is plotted as a function
of time. The steady state strain of 1.4 is reached after a delamination over a length
equal to the interlayer thickness 2 mm.

117
 12

10

Stress σ11 (MPa)

8

0
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)

Figure 7.15: The steady state value reached by the stress on the right end of the
polymer as a function of the crack tip position and the corresponding time.

7.5 Two zones of dissipation

A steady state has been observed as in the TCT experiments. In this steady state
regime, the different zones of stretching of the interlayer will now be studied. The
point of this model is indeed to get a better understanding of the process in the
delamination front vicinity.
Two zones of deformation can clearly be identified in Figure 7.16 where the strain
field in direction 1 (the applied displacement direction) is displayed.
The fast stretching zone appears with a strong gradient in the strain in direction
1. In particular on the median plane of symmetry this strain gradient corresponds to
the one measured in the DIC experiment on the plane where ink dots were sprayed.
In the vicinity of the delamination front, a contraction zone in direction 1 (cor-
responding to a stretching zone in direction 2) is confined in a smaller area. This
second stretching zone corresponds to the expected dissipation zone near the crack
tip.

7.5.1 The fast stretching zone

First, in the fast stretching zone, we will compare our results to the ones obtained
previously from the DIC experiments (see 4.4.2 and 5.3)
On the median axis of symmetry, the strain in direction 1 is plotted in Figure
7.17 as a function of the distance to the point right above the delamination front.

118
 Median axis of symmetry
Stretching in direction 1

NE, NE11

1.56
0.80
0.76
0.71
0.67
0.63
0.58
0.54
0.50
0.45
0.41
0.37
0.32
0.28
0.24
0.19
0.15
0.11
0.06
0.02
−0.02
−0.07
−0.11
−0.15
−0.20
−0.24

Delamination front vicinity

Compression in direction 1

Figure 7.16: The finite element mesh is displayed with the strain in direction 1 field.

In this figure, the strain in direction 1 presents a spatial distribution which is highly
similar to the one observed in the TCT test. Indeed, the strain increases rapidly
over a length equal to the interlayer thickness h and then stabilizes at a steady state
value. Moreover, the strain increases up to 70% of the steady state value on this
distance h which is also similar to the result found in the DIC measurements (see
Figure 4.12). In the experiment, the mean constant strain value 0TCT is reached
much faster than in the finite element model. Indeed, the finite element calculation
is made in real plain strain whereas in the experiment, the width narrowing spans
over a much longer distance (close to the initial width 50 mm). This intermediate
length scale is hidden in these two dimensional calculations.

Similarly to what has been found during the experiments, the strain rate on this
length can be evaluated by ˙11 = ∗ 11 = 7 ∗ 2 ∗ 1.2 =4.2 s−1 . In this zone, the
Vf ront
h
strain rate can reach a higher value as displayed in Figure 7.18. The maximal strain
rate is about 5 s−1 . The strain rates obtained here correspond to the maximum of
dissipation of the viscoelastic interlayer with a relaxation time of 0.01 s.

119
 1.5

Bulk Strain 11

1

0.5

h Far Field 11 = 1.4

0
-5 0 5 10 15 20
Distance to front (mm)

Figure 7.17: Steady state regime strain in direction 1 along the median axis of
symmetry as a function of the distance to the crack tip. Deformation occurs mainly
on a distance equal to the interlayer thickness.

5
Far Field 11 = 1.4

4
Bulk ˙11 (s−1 )

0
-5 0 5 10 15 20
Distance to front (mm)

Figure 7.18: The local steady state strain rate in direction 1 along the median axis
of symmetry as a function of the distance to the crack tip. The mean strain rate is
about 1.5 s−1 but the maximal strain rate is about 5 s−1 .

120
7.5.2 Near crack process zone
In the previous parts, the results were similar to the one found in the TCT experi-
ments. The vicinity of the delamination front will be now described.
The results presented here, have to be taken with caution. Indeed close to
the delamination front, the strain rate and the strain levels encountered by the
interlayer in the real experiment are not properly described by the simple interlayer
behavior used in this finite element model. Moreover, the mesh size might not be
fine enough to describe the crack tip properly. In the vicinity of the crack tip, the
strain distribution is completely different from the one just described in the bulk
material. Indeed, the strain variations are concentrated in a small area around the
crack tip. In this zone, the polymer is in tension in direction 2 and compression
in direction 1. In this zone, the maximal strain in direction 2 is about 0.3. The
large values of strain above 0.25 are concentrated on a surface of about 7.10−4 mm2 .
However, the strain values higher than 10% are covering a much larger area which is
approximately a disk of 0.4 mm of diameter. A rough estimation of the strain rate
along the interface gives a value close to 30 s−1 which is already much higher than
in the fast stretching zone. However, the zone of strain concentration might not be
meshed finely enough to get a proper description of the strain and the strain rates
involved could be higher.
In the representations of the deformation field in direction 1 and 2 (Figure 7.19),
the process zone close to the delamination front can spatially be from the fast
stretching zone. It confirms the separation hypothesis made in 5.3 about the two
different sources of dissipation. It is interesting to emphasize however, that this
spatial separation of the two zones of dissipation does not imply the absence of
interactions. These interactions were not studied here and could be at the core of
future work on this topic.

7.6 Near crack work of fracture

One of the remaining questions at this point is how much energy is dissipated near
the delamination front. To answer this question, the energy flow entering this area
can be determined similarly to the Poynting vector used for electromagnetic fields.
We use the approach of Atkinson and Eshelby [57] based on an energy-momentum
tensor to determine the energy flow ΠS entering a given contour C around an area
A surrounding the crack tip. Note that simultaneously Rice [58] used a similar
approach to develop the concept of J-Integral for the case of elastic fracture. For

121
 NE, NE11

1.56
0.20
0.18
0.16
0.15
0.13
0.11
0.09
0.07
0.05
0.04
0.02
−0.00
−0.02
−0.04
−0.06
−0.07
−0.09
−0.11
−0.13
−0.15
−0.17
−0.18
−0.20
−0.22
−0.24

(a) Strain in direction 1






























(b) Strain in direction 2

Figure 7.19: Strain field near the delamination front in the steady state regime.

122
our viscoelastic material, the specific properties of the J-Integral and especially its
path independence do not hold. The energy flow is here defined by:

dui
Z !
ΠS = W δjk − σij mj ds (7.6.1)
C dxk
As a energy flow, ΠS is a vector. In this equation W is the stored energy density
defined as W = σij dij . σij and ij are the stress and strain field components.
R ij
0
mj is the vector normal to the contour C and ds the curvilinear abscissa along the
same contour. Further information about the different terms and explanation about
this integral can be found in Bower’s book [59].
This integral definition is valid in a region where the strains are small (though
it can be extended to large strain). This is really appropriate in this case as there
is a region in between the fast stretching zone and the delamination front process
zone in which strain is low. In this region, a circular contour C will be drawn and
used to calculate the energy flow. This contour is defined as a circle centered on the
delamination front with a radius defined by the maximal strain encountered on his
path. An example of contour is given in Figure 7.20. Through the interface part of
the contour, the energy flow is equal to Γ∗s . On the open crack part of the contour,
the energy flow is zero. Thus the calculation is only made on the circular part of
the contour. The radius of the circle is defined so it passes through a region where
the maximal strain is lower than 5%.

NE, NE22

0.32
0.30
0.29

Circular contour
0.27
0.25
0.24
0.22
0.20
0.19
0.17
0.15
0.14
0.12
0.10
0.09
0.07
0.05
0.04
0.02
0.00
−0.01

C
−0.03
−0.05
−0.06
−0.08
−0.61

Crack face

Interface
Figure 7.20: Contour used for the energy flow calculation, passing through a “small”
strain area around the process zone in the vicinity of the delamination front.

The delamination propagates in direction k = 1. So to calculate the energy flow

123
we have to determine two terms. The first one corresponds to the work of external
forces done on the contour C which in our case will be approximated by:

1
W δj1 mj = (σ11 11 + σ22 22 + 2σ12 12 )m1 (7.6.2)
2
The second term arises from the rate of change of mechanical energy inside the
area A. As σ and  are symmetric tensors this term can be written as:

dui
σij mj = (σ11 11 + σ12 12 )m1 + (σ22 12 + σ12 11 )m2 (7.6.3)
dx1
In the previously defined loading conditions and material description, the calcu-
lation of ΠS in direction k = 1 gives Π1S ≈ 1 kJ m−2 .
ΠS includes the viscoelastic losses in the interlayer and the viscous and damage
dissipation in the cohesive elements:

ΠS = Γcrack + Γ∗s (7.6.4)

Thus, as Γ∗s = 0.17 kJ m−2 , Γcrack ≈ 0.8 kJ m−2 are dissipated in the interlayer
process zone close to the delamination front. This also means that the energy dissi-
pated by viscoelastic losses in the interlayer stretching can be divided in 2.4 kJ m−2
dissipated in the fast stretching zone and the rest corresponds to the near crack
work of fracture. The ratio between these two kinds of dissipation is comparable to
the one found in the TCT experiment between Γcrack and Πbulk h close to 0.5.
To sum up the result of this paragraph, the different energies are displayed in
Figure 7.21 as parts of the macroscopic work of fracture. A schematic view of the
delaminating interlayer gives the location of the related dissipation zone.

7.7 Impact of interlayer relaxation time and work

of separation
The results found in the previous paragraph depend on both the interlayer behavior
and work of separation. In order to compare their effects on the delamination
process, we will focus on the propagation velocity of the delamination front, the
maximal strain in direction 1 far from the delamination front, the total work, the
energy dissipated in the vicinity of the delamination front and the length of the fast
stretching. For each case, Γ∗s will be evaluated.
The adhesion and the interlayer behavior were also studied during the experi-

124
 (a) Partition of Gm

(b) Location of the dissipation zones

Figure 7.21: The macroscopic work of fracture is dissipated in different zones before
some energy can be delivered to the crack tip for the crack to advance.

125
mental part. Experimentally it is challenging to change one without affecting the
other (see 6). In the finite element model it is easier to change the interlayer behav-
ior without acting on the adhesion and vice-versa. It has to be noted that a shift in
the relaxation time of the interlayer viscoelastic response is equivalent to a change
in the applied velocity.

7.7.1 Work of separation

Three different levels of the effective work of separation were tested numerically.
The three parameters of the cohesive zone were changed simultaneously as it was
found that keeping a similar shape of the traction separation law was leading to less
convergence problem. Thus the standard adhesion refers to the previously used co-
hesive zone set of parameters Γs = 0.038 MPa mm, T 0 =10 MPa and K 0 =7.5 MPa.
The lower and higher levels of adhesion were obtained by respectively dividing and
multiplying the three parameters by a factor 2.
The effective work of separation and effective maximal cohesive stress found for
these three different adhesion levels are given in Table 7.2. The cohesive element
viscosity is unchanged and still equal to η =0.001 s−1 . As one can see, these are
effectively three different levels of adhesion. Due to the viscosity, the effective work
of separation Γ∗s is 4 to 6 times larger than the original Γs and T ∗ is about 1.5 times
T.

Adhesion level Γ∗s T∗ K∗

(kJ m−2 ) (MPa) (MPa mm)
Low ÷2 0.069 7.5 3.75
Standard 0.17 16 7.5
High ×2 0.46 27 15
Table 7.2: Effective cohesive zone parameters for the three different level of adhesion.

In all these simulations, the loading velocity δ̇ is equal to 10 mm s−1 . The change
in adhesion leads to different front propagation velocities. In Figure 7.22, the delam-
ination front position is plotted as a function of time for the three levels of adhesion.
In the three cases, a steady state delamination regime is obtained. The higher the
adhesion, the lower the delamination velocity: 10 mm s−1 for the lowest adhesion,
7 mm s−1 for the standard adhesion, 5.5 mm s−1 for the highest adhesion. Note that
the front is advancing in the direction opposite to the pulling direction. Thus the
stretching of the interlayer is given by: TCT = δ/a.
Figure 7.23 displays the strain 11 calculated along the middle plane of the inter-

126
 Delamination front position (mm)
10

2 Low Adhesion
Standard Adhesion
High Adhesion
0
0 0.5 1 1.5
Time (s)

Figure 7.22: Delamination front position as a function of time for the three levels of
adhesion tested. The delamination reaches a steady state in which the delamination
front velocity is decreasing as the adhesion increases.

layer. This strain increases when adhesion increases as the front velocity decreases
and the loading velocity is constant. This is similar to what has been found in the
experiment: the OTES treatment lead to a weaker interface, the delamination front
velocity was higher and the strain was found to be lower than for the non treated
sample (see Figure 6.5).
For the three levels of adhesion, the fast stretching zone spans over the same
length equal to the interlayer thickness h. This confirms that this effect is solely due
to the sample geometry namely the interlayer thickness. However, the increasing
strain and the decreasing delamination velocity have antagonist effects on the dis-
sipated energy. In the case of the high adhesion, the delamination velocity is about
5.5 mm s−1 and the strain reached over the fast stretching zone is about 1.5. This
leads to a strain rate close to 4.1 s−1 . For the lowest adhesion, the delamination
velocity close to 10 mm s−1 is higher but the strain reached over the fast stretching
zone about 0.7 is lower. This results in a lower strain rate of 3.5 s−1 . Thus a lower
level of adhesion leads to both a lower strain and a lower strain rate which should
lead to a lower amount of dissipated energy from the stretch mechanisms.
This last conclusion is confirmed by the macroscopic work of fracture Gm found
in the steady state regime. In the case of the standard adhesion, Gm is equal to
5.9 kJ m−2 . For the lowest adhesion level, Gm = 2.5 kJ m−2 whereas for the highest
level of adhesion Gm =13 kJ m−2 . This work is partially stored elastically (Γel
bulk ) or

127
 2
High Adhesion
Standard Adhesion
1.5 Low Adhesion

Bulk Strain 11

1

0.5

h
0
-10 -5 0 5 10
Distance to front (mm)

Figure 7.23: Bulk strain measured along the middle plane of symmetry in direction
1 for the three different levels of adhesion. As the adhesion increases the interlayer
is stretched at a higher level of strain.

dissipated (Γvisco
bulk ) in the bulk interlayer: Γbulk = Γbulk + Γbulk .
el visco

It is also dissipated near the crack tip Γcrack . In Figure 7.24, the partition of
this total work is displayed for the three levels of adhesion. The values of each
energy release rate are also given in Table 7.3. As adhesion increases both values of
the energy dissipated near the crack tip and in the bulk of the interlayer increases.
Moreover, the ratio between the near crack work of fracture and the bulk stretch-
ing work is almost constant between the low and standard adhesion (close to 0.2)
whereas it is tripled for the high level of adhesion (about 0.6).

Γbulk Γcrack
Adhesion level Gm Γ∗s
Γel
bulk Γvisco
bulk = J − Γ∗s
Low 2.5 1.3 0.9 0.2 0.067
Standard 5.9 2.5 2.6 0.6 0.17
High 13 2.4 6.4 3.7 0.46
Table 7.3: Different energies dissipated during the steady state delamination for
three different levels of adhesion. All values are in kJ m−2 .

This shows that a change in adhesion affects the whole delamination process from
the interlayer maximal stretch to the strain rates involved. This impacts the different
zones of dissipation of energy in a non proportional manner. It emphasizes the
interactions between the near front process zone and the fast stretching zone. These
interactions, in which Γ∗s and T ∗ play different roles, are yet not well understood.

128
Figure 7.24: Partition of the macroscopic work of fracture (Gm ) required for the
steady state delamination. Energy is either stored elastically in the interlayer (Γel
bulk ),
dissipated by viscoelasticity in the bulk of the interlayer (Γvisco
bulk ) or dissipated near
the crack tip (Γcrack ) before it can be used to separate the interface (Γs ).
∗

7.7.2 Viscoelastic relaxation time

Three different relaxation times τ1 were tested 0.001 s, 0.01 s and 0.1 s. The rest of
the material parameters are kept constant. The pulling velocity is also kept equal
to 10 mm s−1 . The relaxation times have to be compared to this velocity. In Figure
7.25, a calculated cyclic uniaxial tension response up to 200% is plotted for the three
different times at a strain rate of 100 s−1 . This numerical experiment qualitatively
emulates the experiments conducted in Figure 3.12. One can see that a higher
relaxation time is equivalent to a faster pulling speed and a lower relaxation time is
equivalent to a slower pulling speed. The parameters of the cohesive zone for these
three cases are given in Table 7.4. The effective adhesion energy Γ∗s is similar for
the lowest and intermediate times but it is slightly higher for the highest time, due
to the higher crack velocity. The critical stress Tc0 has an effective value which is
similar for the three times. The adhesion will thus be considered as not changed
between the two lower times but significantly different for the highest time. This
difference will be explained in the following paragraphs.
For τ1 =0.001 s and τ1 =0.01 s the delamination reaches a steady state regime
(Figure 7.26). The front velocity is equal to 7 mm s−1 in both cases. The strains are
comparable in the two cases and equal to 1.3 for τ1 =0.001 s and to 1.4 for τ1 =0.01 s
(Figure 7.27). For these two cases, most of the stretch of the interlayer occurs over
a length equal to its thickness.

129
 200
τ1
0.1 s
150 0.01 s
True Stress (MPa)

0.001 s
100

50

-50
0 0.5 1 1.5 2
Nominal Strain
Figure 7.25: Uniaxial tension response at 100 s−1 for the three different relaxation
times.

τ1 Γ∗s T∗ K∗
(kJ m−2 ) (MPa) (MPa mm)
0.001 0.22 18 7.5
0.01 0.17 16 7.5
0.1 0.39 20 7.5
Table 7.4: Effective cohesive zone parameters for the three different times.

130
 On the contrary, for τ1 =0.1 s no steady state is observed. The delamination front
propagation is fast and only slows down because of the viscosity of the cohesive zone
elements. In this case, the strain does not reach a steady state value and is lower
than 0.9 when the calculation ends.

Delamination front position (mm)

8

2 τ1 = 0.001s
τ1 = 0.01s
τ1 = 0.1s
0
0 0.5 1
Time (s)

Figure 7.26: The delamination front position as a function of time for three relax-
ation times τ1 . For τ1 =0.1 s no steady state was reached.

The partition of the macroscopic work of fracture is displayed in Figure 7.28 and
the different energies are summarized in Table 7.5 for the three different relaxation
times.
In the case of τ1 =0.1 s, fast delamination and lower strain lead to a near crack
work of fracture close to 0. Indeed, the ΠS calculation gives that all the energy
flowing to an area close to the delamination front is used to separate the cohesive
elements. None of this energy is dissipated in the vicinity of the crack tip. This could
be due to very large strain rates, so the interlayer responds as an elastic material.
For the two lower times, Γs is equal to 0.2 kJ m−2 . Using a shorter relaxation
time to describe the viscoelasticity of the material is equivalent to a slower pulling
speed. This induces a smaller macroscopic work of fracture. Close to the crack tip,
a larger amount of energy is dissipated for τ1 = 0.001 s−1 . This seems to be due to
a huge maximal strain close to 1.4 reached in this zone. On the contrary, in the
bulk interlayer, lower strain rates and a slightly lower strain, lead to smaller elastic
and dissipated energy compared to the intermediate time. The ratio of viscoelastic
dissipation over the elastic energy is constant close to 0.5 for τ1 = 0.001 s and
τ1 = 0.01 s. Finally, between τ1 =0.001 s and τ1 =0.01 s, the decrease in the near

131
 1.5
τ1 = 0.001 s
τ1 = 0.01 s
τ1 = 0.1 s

Bulk Strain 11

1

0.5

h
0
-10 -5 0 5 10 15
Distance to front (mm)

Figure 7.27: The strain along the mid plane is plotted as a function to the distance
to the point at the vertical of the delamination front for the three relaxation times
τ1 .

crack work of fracture is compensated by the increase in the bulkd stretching work
leading to similar macroscopic work of fracture in the two cases. It appears that the
increase of the relaxation times induces a transfer of the dissipation from the near
crack process zone into the bulk stretching zone.

τ1 (s) Gm Γel
bulk Γvisco
bulk Γcrack Γ∗s
0.001 4.8 0.95 1.3 2.2 0.22
0.01 5.9 2.5 2.6 0.6 0.17
0.1 3.9 0.73 2.8 0 0.39

Table 7.5: Different energies dissipated during the steady state delamination for
three different viscoelastic relaxation times τ1 . All values are in kJ m−2 .

7.8 Coupling between the near crack and bulk

stretch responses.
The previous section shows that the energy dissipation in the near crack process
zone and in the bulk of the interlayer are connected and that this relationship
depends on the material relaxation time and on adhesion. Here we will describe
these interactions using the material behavior.
In [41], Barthel and Fretigny estimate the near crack work of fracture within a
linear viscoelastic theory. In this approach, the creep function of the material is

132
Figure 7.28: Partition of the total work required for the steady state delamination
for three different viscoelastic relaxation times τ1 .

used to connect the far fireld to the near crack field.

This model defines, the length of the cohesive zone dCZ by:

dCZ (tCZ ) = Vf ront tCZ (7.8.1)

where Vf ront is the delamination front velocity and tCZ is the time it takes to the
delamination front to cross the length dCZ .
φ is an effective creep function of the material for an advancing crack. Denoting
k as the ratio between the long term and instantaneous moduli: k = E∞
E0
and for a
Zener model (single relaxation time τ1 , which corresponds to a creep time τc = τ1 /k)
φ can be expressed as:

2 1−k
" #
φ(t) = 1+2 (1 − t/τc − exp(−t/τc )) (7.8.2)
E∞ (t/τc )2
The work of separation is a constant defined as:

π ∗ 2 π
Γ∗s = (T ) φ(0)dCZ (0) = (T ∗ )2 φ(tCZ )dCZ (tCZ ) (7.8.3)
8 8
Finally, the ratio between the near crack work of fracture and the work of sepa-
ration is given by:

Γcrack φ∞
= . (7.8.4)
Γs∗ φ(tCZ )
This last equation shows that the near crack work of fracture is related to the

133
work of separation through the ratio between the behavior of the relaxed mate-
rial (far from the crack) and the material behavior close to the crack. Thus the
underlying hypothesis is that far from the crack the material is in a relaxed state.
This approach will now be used to describe the previous results.
For the different adhesion levels used in 7.7.1, we can retrieve the near crack work
of fracture using the creep function of equation 7.8.2. This method gives results
(Table 7.6) of the same order of magnitude than the one found with the energy flow
ΠS (see Table 7.3). This method also provides an estimation of dCZ and tCZ . The
distance dCZ decreases from 5 to 3 cohesive elements affected by the propagation of
the crack. This is similar to what is found in the finite element calculations. The
frequency at which the material is loaded at the crack tip increases with adhesion.
These frequencies are on the glassy side of the glass transition (Figure 7.7). This
is coherent with the fact that Γcrack is 10 times higher than Γs ∗. According to this
model, the dissipation at the crack tip is saturated for the three levels of adhesion.
Adhesion level dCZ (mm) tCZ (ms) ωCZ = 2π/tCZ (Hz) Γcrack (kJ m−2 ) Γ∗s (kJ m−2 )
Low 0.041 4.1 1530 0.60 0.067
Standard 0.023 3.3 1900 1.54 0.17
High 0.023 4.1 1530 4.1 0.46

Table 7.6: Evaluation of the near crack work of fracture Γcrack and of the frequency
ωCZ in the vicinity of the delamination front using the creep function for three
different levels of adhesion.

The same approach can be taken for the three relaxation times. The near crack
work of fracture and tCZ are evaluated in Table 7.7. The Γcrack values found here
are not comparable to the ones found previously with ΠS (Table 7.5). This can be
explained by the fact that the ratio k used to estimate the creep function φ is no
longer appropriate as for τ1 = 0.001 s and 0.1 s the assumption that the material is
fully relaxed in the far field no longer applies.
τ1 dCZ (mm) tCZ (ms) ωCZ = 2π/tCZ (Hz) Γcrack (kJ m−2 ) Γ∗s (kJ m−2 )
0.001 0.016 2.3 2700 1.33 0.22
0.01 0.023 3.3 1900 1.54 0.17
0.1 0.040 1.0 6300 3.89 0.39

Table 7.7: Evaluation of the near crack work of fracture Γcrack and of the frequency
ωCZ in the vicinity of the delamination front using the creep function for three
different relaxation times.

Thus an extension of this model which could take into account the real behavior
of the interlayer far from the delamination front, would be able to describe properly
this situation.

134
7.9 Conclusion
The delamination of the interlayer has been modeled and some of the most important
features of the experiment were recovered: especially the fast stretching zone length
and the distribution of strain are found to be similar to the ones found in the
experiment. The finite element model also confirmed the separation of the dissipated
energy in two zones of dissipation: a fast stretching zone and a process zone in the
delamination front vicinity. By calculating the energy flow ΠS , the energy dissipated
in this latter zone has been evaluated.
One can also note that the fast stretching zone can be modeled as in paragraph
5.6 by a uniaxial stretching of the interlayer at the appropriate strain rate. Here,
using the polymer model defined in 7.3, at 4.2 s−1 , the interlayer stretched up to
120% dissipates 4.2 kJ m−2 . This value is close to the one deduced from the ΠS
calculation (2.4 kJ m−2 ).
At the crack tip the delamination of the interlayer occurs mainly in mode I
despite the remote loading which is similar to a 0◦ peel test. The blunting of the
interlayer close to the delamination front leads to its stretching in direction 2. In
this near crack process zone the strain rates are of the range from 100 s to 1000 s for
which dissipation is at a maximum.
Finally, the dependence of the delamination response with the interlayer vis-
coelastic relaxation time and work of separation has evidenced strong coupling be-
tween the different zones of dissipation. It was difficult to show clear tendencies
for a change of the viscoelastic time. This is due to the fact that the interlayer
viscoelastic description with only one relaxation time leads to a narrow glass tran-
sition. Changing adhesion had the expected impact on delamination: an increase
in adhesion leads to an increase in interlayer stretching and to an increase of the
dissipated energy. However, a too stiff increase in adhesion should lead to stretch
the interlayer at strain too large for the calculation to converge. Similarly a very
weak adhesion will not be possible in terms of calculation convergence as the front
will propagate “instantaneously“ and a problem similar to the high relaxation time
will be encountered.
Future work will have to focus on the description of the interlayer behavior to
take into account a larger amount of dissipated energy at large strains and high
strain rates. Most of the calculation time is spent on reaching the steady state
delamination. Thus a clear improvement will be to develop an Eulerian formulation
instead of the actual Lagrangian formulation, as described by Dean and Hutchinson
in [60].

135
Main results
• The steady state delamination is recovered:

– The steady state delamination velocity, strain and stresses are

comparable to the experiment.
– The fast stretching zone length is equal to the interlayer thick-
ness.

• The separation of the macroscopic work of fracture in two dissipation

zones has been confirmed.

• The near crack work of fracture is evaluated thanks to the energy flow
ΠS and the different components of the macroscopic work of fracture
have been deduced from there.

• The effects of different levels of adhesion and of different relaxation

times on the delamination response have been studied.

• Using the material creep law, the coupling between the near crack
instantaneous response and the relaxed response of the interlayer, far
from the delamination front, has been displayed.

136
Chapter 8

Conclusions and perspectives

The main purpose of this study was to make the link between impact performance
of laminated glass, interlayer rheology and loss of adhesion. For thin enough glasses,
the impact results in glass breakage, followed by interlayer delamination and stretch-
ing to large deformation. To identify the source of the dissipation mechanisms, the
interlayer mechanical behavior has been thoroughly studied in relation to microstruc-
ture. To connect this behavior to the delamination processes and to characterize
different dissipation zones, a model experiment has been setup and analyzed.

The small strain behavior of the PVB interlayer displays a classical time/tem-
perature dependence (section 3.3.1). At large strain the PVB response is highly
non linear with a complex behavior depending on the applied strain rates and tem-
peratures (section 3.3.2). At high strain rates or low temperatures, the interlayer
presents a yield stress at moderate strains and a strong strain hardening for higher
strains. In this regime more than 80% of the energy provided to stretch the inter-
layer is dissipated. For lower strain rates and higher temperatures these features
disappear and a more classical non linear viscoelastic behavior is observed. About
40% of the energy is dissipated in these conditions.
Using photoelastic measurements (section 3.4) and complementary DSC (section
3.5.1) and SAXS experiments (section 3.5.2), we were able to establish a model for
the interlayer polymer network (3.5.3). In this model, two interpenetrated networks
correspond to two modes of energy dissipation. A first network presents a visco-
plastic like behavior. It is composed of chains crosslinked by hydrogen bond nodules.
The second network is not involved in the H bond nodules, but it is connected to the
rest of the polymer network through entanglements and isolated hydrogen bonds.
Due to the presence of the second viscoelastic network, the plastic response of the
first network can be activated only at high strain rates (section 3.6).

137
 The interlayer behavior being more clearly understood, the question was: What
type of dynamic loading does the polymer undergo during delamination? In the
TCT test, a steady state delamination regime was first identified in a certain range
of temperatures and applied velocities (section 4.2). In this regime, the delamination
occurs in a symmetric way, the delamination fronts are straight and propagate at
constant velocity. The delamination velocity has been found to be proportional
to the applied velocity. The steady state regime is bounded by two other regimes
(section 4.3). Most of the time, at higher velocities or lower temperatures, the
delamination stops due to the rupture of the more brittle interlayer, which is more
sensitive to experimental defects. At lower velocities or higher temperatures an
interesting unstable regime was observed; the delamination fronts start to undulate
and often stops leading to interlayer rupture.
In the steady state delamination regime, two different dissipation mechanisms
were identified (section 5.3). In the immediate vicinity of the delamination fronts
the rupture of the interface combined with a local deformation of the interlayer leads
to a non negligible near crack work of fracture. Ahead of the fronts, the interlayer
gets deformed through its whole thickness in a fast stretching zone at high levels of
strain. Digital image correlation and photoelastic measurements have shown that
this stretching of the bulk material occurs over a length equal to the interlayer
thickness. This result is of prime interest as it enables to estimate the strain rate
applied on the interlayer knowing the delamination velocity. In this zone, interlayer
stretching can be described mainly as a uniaxial tension (section 5.6). Thus we have
demonstrated that a uniaxial tension test, at the appropriate strain rate, can be
used to estimate the volume density of the bulk stretching work.
Finally, the effects of applied velocity, temperature and interfacial toughness on
these dissipation mechanisms have been studied. The near crack work of fracture
and volume density of the bulk stretching work increase when the velocity increases
or when the temperature decreases (section 5.7). Three levels of adhesion have also
been obtained with silanized glass surfaces (section 6.2). At higher level of adhesion,
no delamination was observed and the interlayer was immediately ruptured. For
lower adhesion, the delamination occurs faster and induces a lower strain level than
for the reference level of adhesion (section 6.3.2).

A finite element model has been implemented to describe the delamination

mechanisms in a qualitative manner. The interlayer is described as a viscoelastic-
hyperelastic material and the interface is modeled with cohesive elements. Some
characteristics of the TCT test have been retrieved (section 7.4): the steady state

138
delamination, the large interlayer stretch and the length of the fast stretching zone in
which most of the interlayer deformation occurs as in the experiment. This model
confirms the tentative breakdown of dissipation into two process zones and two
mechanisms. A strong coupling between these two processes appear, but the major
parameters and how they affect the overall response have not been demonstrated
yet. A preliminary estimation of the energy dissipated in the crack vicinity has also
been performed (section 7.6). It has been used to compare different simulations with
different loading conditions and different adhesion levels (section 7.7).

From this picture of the delamination and of the associated dissipative processes
in the TCT test, we can derive a better understanding of the coupling between the
remote loading and the crack tip mechanisms occurring during delamination of this
viscous interlayer. In our case, the polymer ligament is stretched at high levels of
deformation and because of the dissipation mechanisms in the interlayer rheology,
a large amount of energy is dissipated. Similarly to what has been observed in the
peeling of adhesives [61], a large deformation zone is observed ahead of the crack tip.
In the case of thin films of adhesive, the formation of fibrils and their subsequent
rupture leads to the dissipation of energy. However, a second dissipation zone also
exists in the immediate vicinity of the crack, similar to what is observed in thicker
films.

Still, many points of our study could be investigated in future studies.

First, concerning the structure of the PVB interlayer, the size of the H bond
nodules and the distance between these nodules has not been clearly identified by
the SAXS experiments. In particular, the distance in between the nodules could be
better measured by diffraction of visible or near UV light as it seems to be of the
order of hundreds of nm. Moreover, such experiments could confirm the nature of
these nodules.
PVB is not the only interlayer that is being used in the laminated glass industry.
Thus the results that have been established with PVB in the TCT test have to be
compared to other interlayers such as PU which is used in more high end applica-
tions (plane or bulletproof windshields) or EVA which is used in less demanding
applications in terms of impact performance (interior walls, decorative panels,. . . ).

139
The interest of such studies could be to have materials with a simpler mechanical
behavior to further challenge our conclusions.
Concerning the TCT test, a lot of work still needs to be done on the dependence
of the bulk and crack related dissipation processes on different parameters such as
applied velocity and temperature (see 5.7). The trends that have been established
are not as precise as they should be to estimate the energy dissipated during the
impact for different impacting object velocities and at different temperatures. We
believe that this could be an interesting tool from the industrial point of view.
An interesting preliminary work has also been done together with Raphaëlle
Kulis on the dependence of the TCT response on adhesion. There is plenty to be
done on this road. A finer tuning of adhesion could confirm the trends observed
during this study, though it could be difficult to get a higher level of adhesion in
which the polymer does not rupture. We believe that a solution could be to increase
the temperature or to slow down the applied displacement to get a steady state
delamination.
Concerning the finite element modeling a future step could be to transfer most of
the viscous dissipation of the cohesive elements either in the damage dissipation of
these elements or in the interlayer material itself. The transfer of this dissipation will
ensure more control over the interfacial rupture energy and stresses. The transfer in
the damage dissipation of the cohesive elements is not difficult to implement but it
will considerably increase the calculation time. To transfer the viscous dissipation
in the interlayer behavior will be more difficult and requires a better description of
the PVB rheology. A better description of the interlayer behavior is also required to
get a better picture of the processes in the neighborhood of the delamination front.
This description will have to include the plastic like behavior of the PVB at high
strain rates and low temperatures. A Bergström-Boyce model [25] is a good example
of the kind of model that could be implemented to model the interlayer behavior as
it takes into account both large strain and plastic behavior. This model can also be
extended to viscous behavior. The drawback of such a model is the large number of
parameters that have to be tuned.
Up to now, the TCT test model has reproduced the whole initiation and propa-
gation process (even if, there was no attempt to reproduce anything real about the
initiation part of the problem). In this Lagrangian formulation, the whole calcula-
tion, is thus used to reach the steady state delamination regime in which a single
step of the calculation is analyzed. A more efficient way to make this calculation
could be to use a Eulerian formulation as described by Dean and Hutchinson in [60]

140
or Landis et al. in [62].
The whole part of the impact concerning the first microseconds after the impact
has not been tackled at all. It should be of high interest to look during these first
instants of the impact at the effect of the interlayer on the propagation of cracks
in the glass. This work could be related to the results of Vandenberghe et al. [63]
in which the number of cracks is connected to the impacting object velocity and to
the interfacial energy required to propagate a crack in glass. It is possible that a
higher apparent interfacial energy will be measured in the case of laminated glass.
As pointed out by Chen et al. [10] the crack propagation velocity should also be
lower in the case of laminated glass. The effect of the interlayer thickness or of other
relevant parameters such as the glass thickness, the temperature or the humidity
rate and the nature of the interlayer could also be investigated.

141
142
Résumé

Les verres feuilletés sont des structures constituées d’un intercalaire polymère pris
entre deux plaques de verres (Figure 8.1). Cet intercalaire améliore les performances
du verre feuilleté en empêchant la perforation par l’objet impactant et en retenant
les éclats de verre. Ces propriétés en font un matériau particulièrement adapté pour
de multiples applications dans le bâtiment (façades, vitrines, sols et plafonds) ou
dans les véhicules (pare-brises d’avions et de voitures ou vitres blindées).

1 − 10mm
0.38 − 1.52mm
Verre
Intercalaire
Verre

Figure 8.1 : Structure d’un verre feuilleté

Ces verres sont soumis à des tests normés qui consistent généralement en un ou
plusieurs impacts sur une large plaque de verre feuilleté, dans des conditions repro-
duisant une situation particulière. Ainsi, dans le cadre de la norme pour l’habitat
EN 356 une une bille de 4 kg est lâchée d’une hauteur de 3 m à trois reprises, sur une
plaque d’environ 1 m2 . Le test est considéré comme réussi si la bille ne traverse pas
la plaque et si aucun éclat trop important n’est projeté. Ces tests mettent en jeu des
phénomènes rapides (rupture du verre et de l’interface) et un profil de fractures com-
plexe. La mise en place de mesures quantitatives est particulièrement difficile dans
ces conditions. On ne peut donc pas, la plupart du temps, utiliser un tel test pour
remonter à la source d’éventuels problèmes ou pour développer de nouveaux pro-
duits. C’est pour cette raison que des études précédentes ont cherché, entre autres,
à relier des tests à l’échelle du laboratoire avec les résultats des tests normés ([5] et
[9]). Cette relation s’est avérée difficile à mettre en œuvre et de nombreuses lacunes
dans la compréhension des mécanismes de dissipation d’énergie lors de l’impact sont
apparues.
Les évènements qui se succèdent au cours de l’impact sont à la fois complexes et

143
rapides. On peut cependant distinguer deux étapes principales (Figure 8.2). Durant
les dix premières microsecondes de l’impact, un profil de fractures en étoile, assez
nettement défini apparaît sous l’objet impactant. La plaque, à cet instant, n’est
pas encore en flexion. Dans les microsecondes qui suivent, les fissures dans le verre
propagent jusqu’au bord de la structure, rebondissent et, combinées à l’effet de la
flexion de la plaque, génèrent le profil de fractures complexe que l’on observe post
mortem. Durant cette deuxième étape, la flexion de la plaque conduit les morceaux
de verres à s’écarter les uns des autres, en étirant entre eux l’intercalaire qui se
décolle. C’est précisément, l’étirement de l’intercalaire qui permet de dissiper une
grande quantité d’énergie lors de l’impact, comme l’a montré entre autres Nourry
[8].

Figure 8.2 : Le déroulement d’un impact peut être décomposé en deux grandes
étapes. Durant les premières micro secondes, la plaque n’est pas encore globalement
en flexion, un profil simple de fissures en étoile apparaît sous l’objet impactant. Puis,
un profil complexe de fissures se développe notamment sous l’effet de la flexion de
la plaque. Cette flexion conduit par ailleurs à l’écartement des morceaux de verre et
à l’étirement de l’intercalaire délaminé.

Dans cette thèse, nous nous sommes attachés à l’étude de cette étape de l’im-
pact durant laquelle l’intercalaire délamine du verre et subit des déformations im-
portantes. Nous avons pour objectif de caractériser les liens entre le com-
portement de l’intercalaire et l’adhésion afin d’expliquer la quantité im-
portante d’énergie dissipée lors de l’impact.
Pour ce faire nous avons abordé le problème en considérant deux axes d’étude.
Nous avons cherché d’une part à caractériser le comportement de l’intercalaire dans
un régime de déformation et de vitesse approprié. D’autre part nous avons utilisé
une expérience qui modélise l’écartement des morceaux de verre durant l’impact.
Cette expérience nous a permis d’étudier les sollicitations imposées à l’intercalaire
lors de sa délamination.

L’intercalaire le plus couramment utilisé dans les verres feuilletés est le Polyvi-
nylbutyral ou PVB. Il s’agit d’un polymère statistique composé de 3 monomères :
vinylacetate (1-2wt%), vinylalcool (18wt%) et vinylbutyral (80wt%). Le matériau,

144
qui n’est pas réticulé chimiquement, contient également 20% en masse de plastifiant.
Les mesures du comportement en petites déformations permettent d’explorer une
large gamme de fréquences et de températures. Elles montrent notamment que l’in-
tercalaire PVB présente une large transition vitreuse autour de 25 ◦C ainsi qu’une
dépendance en temps et températures qui peut être modélisée en utilisant la loi
WLF. Lors de la délamination l’intercalaire est sollicité dans un régime de grandes
déformations et il est donc nécessaire de caractériser le comportement de l’interca-
laire dans cette gamme d’étirement. Pour cela des essais de traction uniaxiale sont
réalisés à différentes températures et taux de déformation. Des essais cycliques (Fi-
gure 8.3) montrent que le PVB dissipe une quantité d’énergie qui représente jusqu’à
90% du travail fourni à basse température (10 ◦C) ou à grande vitesse de déformation
(1 s−1 ). Dans cette gamme de températures et de taux de déformations, le polymère
présente entre 10 et 40% de déformation un plateau de contrainte, ainsi qu’un fort
raidissement au delà de 100% de déformation. A plus basse vitesse (0,001 s−1 ou à
plus haute température 50 ◦C, le PVB dissipe moins d’énergie mais présente une ex-
tensibilité plus grande. On constate également que le comportement viscoélastique
présente un moins fort raidissement et que le plateau de contrainte disparaît.
100
T = 10 ◦
C, ˙ = 1s−1
T = 20 ◦
C, ˙ = 1s−1
80 T = 20 ◦
C, ˙ = 0.1s−1
True Stress (MPa)

T = 20 ◦
C, ˙ = 0.01s−1
60 T = 20 ◦
C, ˙ = 0.001s−1
T = 50 ◦
C, ˙ = 1s−1

40

20

0
0 0.5 1 1.5 2
Nominal Strain
Figure 8.3 : Cycles en traction uniaxiale à différentes températures et taux de
déformation jusqu’à 200% de déformation. Les vitesses de charge et de décharge
sont identiques.

Afin de relier ce comportement à la structure du matériau, des mesures de bi-

réfringences sont réalisées au cours des essais de traction. Pour cela l’échantillon de
PVB, soumis à une traction uniaxiale, est placé entre deux polariseurs croisés. L’in-
tensité de la lumière monochromatique traversant l’ensemble polariseurs/échantillon

145
est mesurée. L’intensité oscille entre des franges sombres et lumineuses qui reflètent
l’orientation des chaines du matériaux et leurs interactions avec la lumière polarisée.
En particulier, on remarque que le nombre de franges est d’autant plus grand que
le taux de déformation est élevé ou que la température est basse. Ces observations
font écho aux différences de comportement présentées précédemment. Mais l’image
d’un polymère dont les chaines seraient non réticulées chimiquement et seulement
enchevêtrées se heurte à d’autres observations. En effet, lors d’un essai de relaxation
(charge puis maintien du niveau de déformation) on observe que, pendant que la
contrainte relaxe rapidement, le niveau de biréfringence reste globalement constant
(Figure 8.4). La relaxation de la contrainte suggère une relaxation du réseau poly-
mère, au contraire le niveau de biréfringence constant invite à penser que les chaines
restent orientées dans la direction de traction. Cette contradiction peut s’expliquer
par la coexistence de deux réseaux interpénétrés.
200
Intensity
True Stress

40
150

True Stress (MPa)

Intensity

100

20

50

0 0
10−1 100 101 102 103
Time (s)

Figure 8.4 : Lors d’un essai de relaxation suivi par biréfringence on constate que
la contrainte relaxe rapidement alors que le niveau de biréfringence varie peu après
la montée en charge.

Cette hypothèse est étayée par différentes observations. Tout d’abord, des expé-
riences de calorimétrie différentielle montrent qu’une perte d’organisation intervient
autour de 70 ◦C (Figure 8.5).
Des mesures de diffraction de rayon X confirment l’existence d’une structure
organisée. Cette structure pourrait s’expliquer, comme suggéré par Schaefer [23] par
la présence de nodules de liaisons hydrogènes. On sait que les groupes hydroxyles
des monomères vinylalcool sont susceptibles de former de telles liaisons entre les
chaines du PVB. Il est probable que localement des amas de liaisons hydrogènes
se forment. Ces nodules plus solides qu’une liaison hydrogène isolée, engendrent

146
 0
1st cycle
2nd cycle
-2 3rd cycle 9h 40◦ C
-9

Cp (mJ/K)
-4

Cp (mJ/K)
-9.5
-6 -10

-8 -10.5
50 75 100
T (◦ C)
-10

-25 0 25 50 75 100
Temperature ( C) ◦

Figure 8.5 : La calorimétrie différentielle à balayage permet entre autre de mettre

en évidence une transition ordre/désordre autour de 70 ◦C. Cette structure n’est
recouvrée partiellement qu’après une attente de 9 h à 40 ◦C.

localement une organisation plus importante des chaines de polymère environnantes.

Enfin les expériences de diffraction permettent d’estimer que la distance entre ces
nodules est supérieure à 300 µm.
Cette description de la structure du PVB permet d’expliquer la grande quantité
d’énergie dissipée lors de la déformation du matériau.

Ayant identifié les mécanismes de dissipation d’énergie dans le comportement de

l’intercalaire, nous nous sommes intéressés à sa délamination du verre. Pour cela
nous avons utilisé une expérience modèle consistant en une traction uniaxiale sur
un verre feuilleté pré-entaillé (“Through Crack Tensile Test” ou TCTT), identique
à celle conduite par Seshadri dans ses travaux [22] (Figure 8.7).
Nous avons tout d’abord établi que dans une certaine gamme de température
et de vitesse, la délamination se fait de manière stationnaire. Dans ce régime, les
fronts de délamination, relativement réguliers, progressent à vitesse constante. L’in-
tercalaire est étiré entre ces fronts. Nous avons montré que de manière similaire
au pelage d’un élastomère depuis un substrat rigide, la force de délamination et
la déformation macroscopique de l’intercalaire atteignent rapidement un plateau.
Notamment, le PVB est étiré à des niveaux de déformation de l’ordre de 150% à
20 ◦C et pour une vitesse de déplacement de 10 mm s−1 (Figure 8.8). Dans ce régime
stationnaire nous avons montré que la délamination de l’intercalaire dépend de la
vitesse de déplacement imposée et de la température.

147
 Liaison H intra

>300nm

Liaison H inter

Nodule de liaisons H
σγ µγ µi
Molécules de plastifiant
µ∞
Chaines impliquées dans les nodules de liaisons H µ0γ
Chaines sans nodules de liaisons H τi
Monomères alcool vinylique porteurs d'un groupe hydroxyle τγ
Liaisons hydrogène

(a) Structure du PVB (b) Modèle rhéologique du

PVB

Figure 8.6 : L’intercalaire présente une structure constituée de deux réseaux inter-
pénétrés.

F δ

Front de
Delamination
a

Figure 8.7 : Vue schématique de l’expérience modèle de délamination sur verre

pré-entaillé : Through Crack Tensile Test.

148
 600 2

Nominal Strain TCT

450 1.5

Force FTCT (N)

300 1
40

δ (mm)
20 0TCT ≈ 1.4
150 0.5
0
0 10 20 30 Force
Strain
0 2a (mm) 0
0 1 2 3 4 5
Time (s)

Figure 8.8 : Force de délamination et déformation de l’intercalaire sont constants

au cours de la délamination stationnaire à 20 ◦C et 10 mm s−1 .

En revanche lorsque la température augmente ou lorsque la vitesse diminue, on

observe un régime de délamination que nous qualifions d’instable. Dans ces condi-
tions, les fronts ondulent et s’arrêtent conduisant à la rupture de l’intercalaire. À
hautes vitesses et à basses températures, un problème différent empêche la mesure
de la force de délamination et de la déformation. En effet, le polymère, plus fragile,
devient plus sensible aux défauts de l’expérience et notamment à ceux introduis par
un mauvais alignement des mors ou par la préparation de l’échantillon. Si la propa-
gation semble d’abord similaire au régime stationnaire de délamination, la rupture
de l’intercalaire met rapidement fin à l’expérience.
Dans le régime stationnaire, il nous est possible de calculer un travail de fracture
appliqué loin du front à partir de la force et de la déformation du PVB. Ce travail, de
l’ordre de 10 kJ m−2 est largement supérieur à l’énergie des liaisons hydrogènes qui
constituent l’interface verre/polymère, de l’ordre de 0,01 kJ m−2 . Cette amplification
de l’énergie dissipée est liée aux grandes déformations que subi le PVB. La Figure
8.9 montre l’évolution de ce travail en fonction de la température et de la vitesse de
déplacement appliquée, ainsi que les différents régimes de délamination.
Afin de déterminer les différentes zones sur lesquelles l’intercalaire est sollicité
lors de la délamination, nous avons montré que le travail de fracture loin du front
est une fonction affine de l’épaisseur de l’intercalaire (Figure 8.10). On peut voir
notamment sur cette figure que l’ordonnée à l’origine est loin d’être négligeable. Ce
résultat suggère que ce travail peut être décomposé en une composante volumique
et une composante surfacique. La partie volumique correspond à l’étirement d’un
ligament de PVB et elle est caractérisée par la pente de la droite précédente (densité

149
 Macroscopic Work of Fracture Gm (kJ/m2 )
35
5◦ C
20◦ C
30
50◦ C

25

20 Brittle
Regime
15
Steady State
10 Delamination

5
Unstable
Delamination
0
10−2 10−1 100 101 102
Velocity δ̇ (mm/s)

Figure 8.9 : Le travail fourni loin du front varie en fonction de la température et de

la vitesse de déplacement imposée. Les limites entre les trois régimes de délamination
sont également grossièrement indiquées sur cette figure.

d’énergie dissipée en volume Πbulk ). La partie surfacique correspond elle, non seule-
ment à la rupture des liaisons hydrogènes à l’interface, mais aussi à la déformation
de l’intercalaire à proximité du front de délamination. Cette dissipation au voisinage
du front est caractérisée par l’ordonnée à l’origine Γcrack .
C’est donc dans deux zones distinctes que le PVB est sollicité. La zone à proxi-
mité du front de délamination est difficilement observable expérimentalement. En
revanche, nous avons pu caractériser la zone de déformation volumique de l’inter-
calaire en utilisant la corrélation digitale d’images. Nous avons pour cela déposé un
mouchetis d’encre au milieu d’un empilement de deux feuilles d’intercalaire assem-
blées dans un verre feuilleté. Ces mesures nous ont permis de déterminer le champ de
déformation sur le plan médian de l’expérience de délamination. Nous avons ainsi pu
observer que l’intercalaire se déforme sur une distance égale à son épaisseur (Figure
8.11). Cette courte distance d’étirement associée au fort niveau de déformation et
à la grande vitesse de délamination, implique des taux de déformation importants
(de l’ordre de 3 s−1 à 20 ◦C pour une vitesse de déplacement imposée de 10 mm s−1 ).
Nous avons alors montré, que la sollicitation dans la zone d’étirement rapide de l’in-
tercalaire pouvait être assimilée à un essai de traction uniaxiale, à une température
et à un taux de déformation appropriés. Ce modèle fournit une méthode simple pour
quantifier la densité d’énergie dissipée en volume par l’étirement du PVB lors de la

150
 Macroscopic Work of Fracture Gm (kJ/m2 )
20

15

10
dGm
dh
= Πbulk
= 8±0.9 MJ.m−3
5

2Γcrack = 4±0.8 kJ.m−2

0
0 0.5 1 1.5
Thickness h (mm)

Figure 8.10 : Le travail fourni loin du front est un fonction affine de l’épaisseur
suggérant qu’il existe deux zones de dissipation d’énergie : l’une où la dissipation
provient de la déformation en volume de l’intercalaire, l’autre en relation avec les
processus à proximité du front de délamination.

délamination.
Afin d’étudier l’impact du niveau d’adhésion sur ces mécanismes dissipatifs, il
faut pouvoir modifier l’interface verre/polymère sans changer le comportement de
ce dernier. Les travaux réalisés avec Raphaëlle Kulis dans le cadre de son stage de
master [13] ont montré la possibilité de modifier l’adhésion en silanisant la surface
du verre. Ainsi, le (3-Aminopropyl) triethoxysilane (APTS) permet d’augmenter
fortement l’adhésion verre/PVB. Au contraire l’utilisation de l’Octyltriethoxysilane
(OTES) la diminue sensiblement. Avec l’APTS, aucune rupture de l’interface n’a
pu être observée en utilisant l’expérience de délamination. Dans le cas de la faible
adhésion, une délamination stationnaire a pu être observée. Cependant le régime
de délamination stationnaire semble être déplacé vers les basses températures et
hautes vitesses en comparaison avec le niveau d’adhésion standard. Les expériences
avec l’OTES, ont aussi montré qu’une adhésion plus faible conduisait, toutes choses
égales par ailleurs, à des niveaux de déformation et de force plus faibles au cours de
la délamination. Ainsi, l’énergie dissipée dans ce cas s’en trouve diminuée.

L’exploration de la délamination de l’intercalaire est approfondie à l’aide d’un

modèle éléments finis. Des éléments cohésifs sont utilisés pour modéliser l’inter-

151
 1
h = 1.52 mm
h = 0.76 mm
0.8

0.6
DIC /0TCT
0.4
A
0.2

-2 -1 0 1 2 3 4
xF /h

Figure 8.11 : La mesure par corrélation digitale d’image de la déformation, dans

la direction de traction et à proximité du front, montre que l’intercalaire est étiré
sur une distance égale à son épaisseur. L’amplitude de la déformation représente
environ 60% de la déformation maximale. On a représenté ici, la déformation locale
normalisée par la déformation loin du front en fonction de la distance à la verticale
du front de délamination normalisée par l’épaisseur de l’intercalaire.

face pendant que l’intercalaire est modélisé comme un polymère viscoélastique non-
linéaire. Cette modélisation partielle du comportement de l’intercalaire, ne permet
pas de reproduire parfaitement l’expérience de délamination mais conduit tout de
même à des résultats qualitatifs intéressants. Ce modèle a tout d’abord permis de
confirmer certains résultats expérimentaux. On retrouve ainsi une délamination sta-
tionnaire, un niveau de force et de déformation loin du front constant et une longueur
de la zone de déformation de l’intercalaire égale à son épaisseur. Le modèle confirme
également la localisation des mécanismes dissipatifs dans deux zones distinctes spa-
tialement : l’une en volume, l’autre à proximité du front de délamination (Figure
8.12). Dans cette deuxième zone, que nous n’avions pas pu explorer expérimen-
talement, l’intercalaire est principalement sollicité dans la direction orthogonale à
l’interface. Ainsi, malgré l’application d’un déplacement en mode II, l’ouverture de
l’interface se fait en mode I en raison de l’émoussement important de la fissure. Nous
avons également pu estimer grâce à ce modèle l’énergie dissipée dans chaque zone.
Enfin nous avons montré l’influence de différent paramètres comme l’adhésion, le
comportement de l’intercalaire ou les conditions de chargement sur les différentes
composantes de l’énergie dissipée. Cela nous a permis de mettre en évidence les re-
lations existantes entre le chargement lointain et les processus locaux de dissipation

152
 NE, NE11

1.56
0.80
0.76
0.71
0.67
0.63
0.58
0.54
0.50
0.45
0.41
0.37
0.32
0.28
0.24
0.19
0.15
0.11
0.06
0.02
−0.02
−0.07
−0.11
−0.15
−0.20
−0.24

Figure 8.12 : Le modèle éléments finis permet entre autre de visualiser deux zones
de dissipation spatiallement distinctes : l’une en pointe de fissure l’autre dans le
volume du matériau. Ici, le champ de déformation dans la direction de traction
(direction 1) présente un gradient positif le long de l’axe médian et dans le volume
du ligament de polymère ainsi qu’un gradient négatif (contraction) au voisinage de
la fissure

d’énergie.

La description de la structure de l’intercalaire et des mécanismes dissipatifs as-

sociés permettent d’imaginer l’utilisation de polymères présentant des liaisons phy-
siques équivalentes à celle du PVB, pour le développement de nouveaux verres feuille-
tés. Par ailleurs l’étude de la délamination de l’intercalaire a permis d’identifier et
de caractériser l’origine de la dissipation d’énergie lors d’un impact sur un verre
feuilleté. Au delà de la volonté de confirmer certaines observations préliminaires,
notamment en ce qui concerne l’influence de la modification de l’interface, les efforts
futurs devront sans doute s’attacher à étudier l’influence de l’intercalaire sur les pre-
miers instants de l’impact. La modélisation éléments finis que nous avons proposée
ici est en outre très incomplète et un raffinement de la description du comportement
de l’intercalaire ainsi que de la pointe de fissure sont envisageables.

153
154
Appendices

A Effect of the thermal treatment during lami-

nated glass preparation on the mechanical be-
havior of the interlayer
Three different PVB were tested.

• Standard PVB as received (referred to as Standard PVB).

• PVB subjected to the heat treatment of the autoclaving process (1,5 h, 130 ◦C
and 10 bar), in-between two pieces of glass put protected by a silicon film from
adhesion to glass (referred to as Thermally treated PVB).

• PVB delaminated from the glass after the laminated glass assembly and the
heat treatment of the autoclaving process (referred to as the Delaminated
PVB).

Two mechanical tests were conducted. For small strain measurements, DMA
experiments were conducted on a range of temperature from −40 ◦C to 60 ◦C (steps of
10 ◦C), for frequencies ranging from 0,1 Hz to 10 Hz. For large strain measurements,
a uniaxial tension test was conducted at 20 ◦C and 1 s−1 strain rate.
There is almost no differences between the standard PVB, the thermally treated
PVB and the delaminated PVB. This also shows that the PVB, even after it has
been subjected to large deformation during a delamination experiment (more than
100% deformation), does not present any plastic deformation and that it’s behavior
is the same after a few hours of recovery.
A small difference can be noted at long time for the DMA results. However, in
the range of corresponding temperature from 50 ◦C to 60 ◦C the DMA starts to loose
its accuracy and this difference is difficult to relate to an actual difference in the
mechanical behavior.

155
 30
Storage Modulus (MPa)

25
2
10 20

σN (MPa)
15
10
Standard 5 Standard
Thermally treated Thermally treated
100 Delaminated 0 Delaminated

-20 0 20 40 60 0 1 2
Temperature (◦ C) N
(a) DMA (b) Uniaxial tension

Figure 8.13: Comparison between the Standard PVB, a cured PVB and a delami-
nated PVB. No large difference was found on the small or large strain behavior.

B Arruda Boyce Model

A polymer chain can be seen as a collection of N monomer blocks of length l. The
chain can then be described by the length r in between its two ends (Figure 8.14).
This distribution of distance r is characterized by its probability density function.
The initial chain length r0 is simply given by a random walk of N steps of length l:
√
r0 = N l

r
y

x
Figure 8.14: Polymer chain in a random position that can be described by different
statistical model. (after [64])

This function can be a Gaussian probability but in this case one has to assume
that the chain is always far from its complete extension rl = N l. In order to describe
the chain extensibility statistic at high strains, a Langevin statistic can be used [64].
However, these model describe a single chain outside of the network. Hence,

156
Flory [65] and Treloar [66] were using a tetrahedral model of the network in which,
three polymer chains are connected together in the center of a tetrahedron on one
end and at the summit of the tetrahedron on the other end. When a deformation is
applied on this small network all the chain will be stretched according to the choosen
probability. However, due to the anisotropic nature of the initial chain orientation,
the system entropy is difficult to calculate as it requires to average the different
orientation contributions. Thus, Arruda and Boyce used a eight chain model (Figure
8.15), which present a cubic symmetry regarding the principal direction of stretch.

Figure 8.15: The eight chain network model used by Arruda and Boyce [16].

We can first define the extension of the chain as r

Nl
which is defined by the
Langevin function:

r/N l = coth β − 1/β (B.1)

β is the inverse Langevin function β = L−1 ( Nr l ). Treloar [66] gives the logarithmic
probability density p(r):
!
r β
ln p(r) = c − N β + ln (B.2)
Nl sinh β
with c a constant. From this probability density, the probability P (r) is obtained
by

P (r)dr = 4πr2 p(r)dr (B.3)

From this the entropy is defined by s = k ln p(r):

!
r β
s = c − kN
0
β+ (B.4)
Nl sinh β
with c0 another constant.
This result is given for a single chain. If we denote n the chain volumic density,

157
as all chain are equivalent, the global entropy is: S = ns. Finally, the Helmholtz free
energy is:
!
r β
W = T ∆S = N kT n β + ln − T c00 (B.5)
Nl sinh β
with c00 an other constant and T the temperature.
Now this model is used inside the eight chain network model. In this network
model, the length of the chain at a given state of stretch is given by:

1 √ 1 √ 1/2
r = √ N l(λ21 + λ22 + λ23 )1/2 = √ N lI1 (B.6)
3 3
where I1 is the first invariant I1 = λ21 + λ22 + λ23 and the λi the stretch in direction
√
i. Moreover, the stretch of the chain is defined as λchain = r/r0 = r/ N l and it can
1/2
now be written λchain = √1 I
3 1
By using equation B.5, one gets:

√
!
β
W = nkT N λchain β + ln − T c00 (B.7)
sinh β
As the Langevin function is not so easy to use and to differentiate to get the
stress expression, most of the time the following approximation is used:

1 1 11
 
W =nkT (I1 − 3) + 2 (I12 − 9) + 3 (I 3 − 27)
2 20N 1050N 2 1
19 519
 
+nkT 4 (I 4
− 81) + 5 (I 5
− 243) (B.8)
7000N 3 1 673750N 4 1
√
We call the maximal chain extensibility λm which is given by λm = N l/ N l =
√
N and µ the shear modulus given by µ = nkT . For a uniaxial traction test in
direction 1 λ1 = λ and for an incompressible material λ1 λ2 λ3 = 1 we get the
following expression of the stress:

dW
σ =λ
dλ
1 X
 5
iCi I1i−1

=2µ λ − 2 . (B.9)
λ i=1 λ2(i−1)
m

Where the Ci are the coefficients of the Langevin function: C1 = 1/2, C2 = 1/20,
C3 = 11/1050, C4 = 19/7000 and C5 = 519/673750.

158
Bibliography

[1] Jean-Marie Michel. Contribution à l’histoire industrielle des polymères en

france : l’histoire du verre triplex, 2012.
[2] European Committe for Standardization. European Standard – EN 12600.
Glass in building - Pendulum test - Impact test method and classification for
flat glass., 2003.
[3] European Committe for Standardization. European Standard – EN 356. Glass
in building - Security glazing - Testing and classification of resistance against
manual attack., 2000.
[4] J.R. Huntsberger. Adhesion of plasticized poly(vinyl butyral) to glass. Journal
of Adhesion, 13:107–129, 1981.
[5] B. Vidal. Modélisation d’impacts sur vitrages feuilletés. PhD thesis, Ecole
supérieure d’Arts et Métiers, 1998.
[6] F.W. Flocker and L.R. Dharani. Stresses in laminated glass subject to low
velocity impact. Eng. Struc., 19(10):851–856, 1997.
[7] P.A. Hooper, R.A.M. Sukhram, B.R.K. Blackman, and J.P. Dear. On the blast
resistance of laminated glass. Int. Journ. Sol. Struc., 49:899–918, 2012.
[8] E. Nourry. Comportement des vitrages feuilletés. PhD thesis, Ecole supérieure
d’Arts et Métiers, 2005.
[9] R. Decourcelle. Comportement mécanique des vitrages feuilletés sous charge-
ments statiques et dynamiques. PhD thesis, Université de Rennes 1, 2011.
[10] J. Chen, J Xu, X. Yao, Liu B, X. Xu, Y. Zhang, and Y. Li. Experimental
investigation on the radial and circular crack propagation of PVB laminated
glass subject to dynamic out-of-plane loading. Engineering Fracture Mechanics,
112–113:26–40, 2013.
[11] J. Chen, X Xu, Yao, X. Xu, B. Liu, and Y. Li. Different driving mechanisms of
in-plane cracrack on two brittle layers of laminated glass. International Journal
of Impac Engineering, 69:80–85, jul 2014.

159
[12] S. Muralidhar, A. Jagota, S.J. Bennison, and S. Saigal. Mechanical behaviour
in tension of cracked glass bridged by an elastomeric ligament. Acta Materialia,
48:4577–4588, 2000.
[13] R. Kulis. Role of the pvb/glass adhesion on the behaviour of laminated glass.
Master’s thesis, École polytechnique féédérale de Lausanne, 2016.
[14] Malcolm L. Williams, Robert F. Landel, and John D. Ferry. The temperature
dependence of relaxation mechanisms in amorphous polymers and other glass-
forming liquids. Journal of the American Chemical Society, 77(14):3701–3707,
1955.
[15] E. Wiechter. Gesetze der elastischen nachwirkung für constante temperatur.
Annalen der Physik, 286(10):335–348, 1893. Annalen der Physick published by
Wiley in 2006.
[16] E.M. Arruda and M.C. Boyce. A three dimensional constitutive model for the
large stretch behavior of rubber elastic materials. Journal of the Mechanics and
Physics of Solids, 41(2):389–412, 1993.
[17] L.R.G. Treloar. The photo-elastic properties of rubber. part i.: Theory of the op-
tical properties of ststrain rubber. Transaction of the Faraday Society, 43:277–
284, 1947.
[18] F. Mertz. Etude de la synthèse et de la caractérisation du Polyvinylbutyral.
Influence des caractéristiques du polymère sur la structure des films. Répartition
de l’Eau au sein des Films plastifiés. PhD thesis, ENSCMu, 1992.
[19] D. Klock. Contribution à l’étude des feuilletés verre-polymère: synthèse et struc-
ture du poly(vinylbutyral) en relation avec ses propriétés mécaniques et adhé-
sives. PhD thesis, ENSCMu, 1989.
[20] P.A. Hooper and B.R.K. Blackman. The mechanical behavior of poly(vynil
butyral) at different strain magnitudes and strain rates. Journal of Material
Science, 47:3564–3576, 2012.
[21] J. Xu, Y. Sun, B. Liu, M. Zhu, X. Yao, Y. Yuan, Y. Li, and C. Xi. Experi-
mental and macroscopic investigation of dynamic crack patterns in PVB lami-
nated glass sheets subject to light-weight impact. Engineering Failure Analysis,
18:1605–1612, 2011.
[22] M. Seshadri. Mechanics of glass-polymer laminates using multi-length scale
cohesive zone Models. PhD thesis, Carnegie Mellon University, Department of
civil and environmental engineering, Carnegie Mellon University, June 2001.

160
[23] J. Schaefer, J.R. Garbow, E.O. Stejskal, and J.A. Lefelar. Plasticization of
poly(butyral-co-vinyl alcohol). Macromolecules, 20(6):1271–1278, 1987.
[24] M. Sonego, L.C. Costa, and J.D. Ambrosio. Flexible thermoplastic composite
of polyvinyl butyral (pvb) and waste of rigid polyurethane foam. Scientific
technical, 25(2):175–180, 2015.
[25] J.S. Bergstrom and M.C. Boyce. Constitutive modeling of the large strain time-
dependent behavior of elastomers. Journal of Mechanics and Physics of Solids,
46(5):931–954, 1998.
[26] I. Lapczyk and J.A. Hurtado. A viscoelastic-elastoplastic finite strain frame-
work for modeling polymers. In ASME 2014 International Mechanical Engi-
neering Congress and Exposition, page 10, 2014.
[27] J. Pelfrene, S. Van Dam, and W. Van Paepegem. Numerical analysis of the
peel test for characterisation of interfacial debonding in laminated glass. Inter-
national Journal of Adhesion and Adhesives, 62:146–153, 2015.
[28] A. Jagota, S.J. Bennison, and C.A. Smith. Analysis of a compressive shear test
for adhesion between elastomeric polymers and rigid substrates. International
Journal of Fracture, 104:105–130, 2000.
[29] M. Seshadri, S. Saigal, A. Jagota, and S.J. Bennison. Scaling fracture energy
in tensile debonding of viscoelastic films. Jour. App. Phys., 101:1–7, 2007.
[30] Y. Sha, C. Y. Hui, E.J. Kramer, P.D. Garret, and J.W. Knapczyk. Analysis of
adhesion and interface debonding laminated safety glass. Journal of adhesion
science and technology, 11:49–63, 1997.
[31] M. Overend, C. Butchart, H. Lambert, and M. Prassas. The mechanical per-
formance of laminated hybrid glass units. Composite Structures, 110:163–173,
2014.
[32] S. Nhamoinesu and M. Overend. Simple models for predicting the post fracture
behaviour of laminated glass. In In the Proceedings of the XXV A.T.I.V 2010,
International Conference, Parma, Italy, 2010.
[33] D. Delincé, D. Sonck, J. Belis, D. Callewaert, and R. Van Impe. Experimental
investigation of the local bridging behaviour of the interlayer in broken lam-
inated glass. In International Symposium on the Application of Architectural
Glass, pages 41–49, Munich, 2008.
[34] S. Malcikan. Numerical Momodell of adhesive failure in delamination of lam-
inated glass. PhD thesis, Gent Universiteit, Faculteit Ingenieurswetenshappen
en Architectuur, 2013.

161
[35] C. Butchart and M. Overend. Delamination in fractured laminated glass. In
International Conference at glasstec, Düsseldorf, Germany, 2012.
[36] D. Ferretti, M. Rossi, and G. Royer-Carfagni. Through crack tensile delamina-
tion tests with photoelastic measurements. In IOS Press, editor, Challenging
Glass 3 - Conference on Architectural and Structural Application of Glass, pages
641–652, 2012.
[37] R.A. Schapery. A theory of crack initiation and growth in viscoelastic media :
I. theoretical development. Int. Journ. of Fracture, 11:141–159, 1975.
[38] P.G. de Gennes. Fracture d’un adhésif faiblement réticulé. C.R. Acad. Sci.
Paris, 307(2):1949–1953, 1988.
[39] F. Saulnier, T. Ondarçuhu, A. Aradian, and E. Raphaël. Adhesion between a
viscoelastic material and a solid surface. Macromolecules, 37:1067–1075, 2004.
[40] A.N. Gent. Adhesion and strength of viscoelastic solids. is there a relationship
between adhesion and bulk properties? Langmuir, 12:4492–4496, 1996.
[41] E. Barthel and C. Fretigny. Adhesive contact of elastomers: effective ad-
hesion energy and creep function. Journal of Physics D: Applied Physics,
42:195302–195311, 2009.
[42] C. Creton and M. Ciccotti. Fracture and adhesion of soft materials: A review.
Rep. Prog. Phys., 79(4):046601, 2016.
[43] B.Z. Newby and M.C. Chaudhury. Effect of interfacial slippage on viscoelastic
adhesion. Langmuir, 13(6):1805–1809, 1997.
[44] N. Amouroux, J. Petit, and L. Leger. Role of interfacial resistance to shear
stress on adhesive peel strength. Langmuir, 17(21):6510–6517, 2001.
[45] B. Cotterell and J.K. Reddel. The essential work of plane stress ductile fracture.
International journal of fracture, 13(3):267–277, 1977.
[46] F.M. Fowkes, D.W. Dwight, J.A. Manson, T.B. Lloyd, D.O. Tichler, and B.A.
Shah. Enhancing mechanical properties of polymer composites by modification
of surface acidity or basicity of fillers. In Materials Research Society, editor,
MRS Proceedings, volume 119, pages 223–235, 1988.
[47] R. Faure and A. Lecointre. Overview of silanes use as a monolayer in sgr.
Technical Report SGR/PCRS – RF/SR – N0293/15, Saint-Gobain Recherche,
2015.
[48] A. Miller and J. Berg. Effect of silane coupling agent adsorbate structure on
adhesion performance with a polymeric matrix. Composites Part A: Applied
Science and Manufacturing, 34(4):327–332, 2003.

162
[49] Y.Y. Lin and H.Y Chen. Effect of large deformation and material nonlinearity
on the JKR (Johnson–Kendall–Roberts) test of soft elastic materials. Journal
of Polymer Science Part B: Polymer Physics, 44(19):2912–2922, 2006.
[50] P. Rahulkumar, A. Jagota, S.J. Bennison, S. Saigal, and S. Muralidhar. Poly-
mer interfacial frature simulation using cohesive elements. Acta Materialia,
47:4161–4169, 1999.
[51] D.S. Dugdale. Yielding of steel sheets containing slits. J. Mech. Phys. Solids.,
8:100–104, 1960.
[52] V. Tvergaard and J. W. Hutchinson. The influence of plasticity on mixed mode
interface toughness. Journal of the mechanics and physics of solids, 41(6):1119–
1135, 1993.
[53] X.P. Xu and A. Needleman. Void nucleation by inclusion debonding in a crys-
tal matrix. Modelling and simulation in materials science and engineeering,
1(2):111–132, 1993.
[54] P.P. Camanho and C.G. Davila. Mixed-mode decohesion finite elements for the
simulation of delamination in composite materials. Technical report, NASA,
2002.
[55] J.F. Faou, G. Parry, S. Grachev, and E. Barthel. Telephone cord buckles –
a relation between wavelength and adhesion. Journal of the mechanics and
physics of solids, 75:93–103, 2015.
[56] Y.F. Gao and A.F. Bower. A simple technique for avoinding convergence prob-
lems in finite element simulations of crack nucleation and growth on cohe-
sive interfaces. Modelling and simulation in materials science and engineering,
12:453–463, 2004.
[57] C. Atkinson and J. D. Eshelby. The flow of energy into the tip of a moving
crack. International Journal of Fracture Mechanics, 1968.
[58] J.R. Rice. A path independent integral and the approximate analysis of strain
concentration by notches and cracks. Journal of Applied Mechanics, 35:379–386,
1968.
[59] A.F. Bower. Applied Mechanics of Solids. CRC Press Taylor and Francis, 2010.
[60] R.H. Dean and J.W. Hutchinson. Quasi-static steady crack growth in small
scale yielding. Fracture Mechanics: 12th Conference ASTM STP, 700:383–405,
1980.
[61] R. Villey, C. Creton, P.P. Cortet, M.J. Dalbe, T. Jet, B. Saintyves, S. Santucci,
L. Vanel, D.J. Yarusso, and M. Ciccotti. Rate dependent elastic hysteresis

163
 during the peeling of pressure sensitive adhesives. Soft Matter, 11(17):3480–
3491, 2015.
[62] C.M. Landis, T. Pardoen, and J.W. Hutchinson. Crack velocity dependent
toughness in rate dependent materials. Mechanics of Materials, 32:663–678,
2000.
[63] N. Vandenberghe, R. Vermorel, and E. Villermaux. Star-shaped crack pattern
of brocken windows. Physical Review Letters, 110:174302–1–5, 2013.
[64] W. Kuhn and F. Grun. Beziehungen zwischen elastischen konstanten und
dehnungsdoppelbrechung hochelastischer stoffe. Kolloid-Zeitschrift, Volume
101(3):248–271, 1942.
[65] P.J. Flory and Jr. Rehner. Statistical mechanics of cross-linked polymer net-
works i. rubberlike elasticity. The journal of chemical physics, 11:512–520, 1943.
[66] L.R.G. Treloar. The elasticity of a network of long-chain molecules iii. Trans-
actions of the Faraday Society, 42:83–94, 1946.

164
165
 Laminated glass: dynamic rupture of adhesion

Laminated glass has been discovered more than a century ago. It is composed of a polymeric interlayer sand-
wiched in-between two glass plies. This interlayer dramatically enhances the performance during impact. Even
if the glass breaks, the partial delamination and the stretching of the interlayer will dissipate a large amount of
energy. This dissipation will protect people from the impacting object while the glass splinters will stick on the
interlayer, preventing harmful splinters projection. During my PhD work, I have identified and characterized
the dissipation mechanisms associated with the interlayer rheology and its delamination from glass.
Using uniaxial traction tests combined with photoelastic measurements, a relationship between the polymer
structure and its mechanical behavior has been provided. The different dissipating mechanisms of the interlayer
rheology have been identified in a rheological model.
In order to understand how the interlayer mechanical behavior is involved during the lost of adhesion at the
glass interface, a model delamination experiment has been setup. This test consists in a uniaxial traction on a
pre-cracked laminated glass sample. In a certain range of applied velocities and temperatures, a steady state
delamination regime has been observed. In these steady state conditions, two zones of dissipation have been
identified. Digital image correlation has been used to quantify the stretching dynamics of the interlayer ahead
of the delamination fronts and to explain the large dissipation observed during the impact. Finally a finite
element model has been developed to confirm experimental observations and to explore the close vicinity of the
delamination fronts.

Keywords: Adhesion, Laminated Glass, Rupture, Interface, Viscoelasticity, Large strain

Verre feuilleté : rupture dynamique d’adhésion

Il y a plus d’un siècle que les verres feuilletés ont été découverts. Il s’agit d’une structure formée d’un intercalaire
polymère pris entre deux plis de verre. Cet intercalaire améliore considérablement les performances à l’impact
de l’assemblage. Ainsi, même si le verre se brise, la délamination et l’étirement de l’intercalaire dissipent une
grande quantité d’énergie. Les personnes sont protégées par cette dissipation de l’objet impactant qui ne traverse
pas le verre feuilleté, mais aussi des projections potentiellement dangereuses d’éclats de verre. Au cours de cette
thèse, nous avons identifié et caractérisé les mécanismes de dissipation d’énergie associés au décollement de
l’intercalaire et à l’étirement qui s’en suit.
Des tests de traction uniaxiale et des mesures de biréfringence ont permis de relier le comportement de l’inter-
calaire à sa structure chimique. Les différents mécanismes dissipatifs du comportement de ce polymère ont été
identifiés et décrit dans un modèle rhéologique.
Nous avons utilisé une expérience modèle afin d’établir les sollicitations subies par l’intercalaire lors de sa
délamination du verre. Cette expérience consiste en un essai de traction uniaxiale sur un verre feuilleté pré-
entaillé. Nous avons montré l’existence d’un régime de délamination stationnaire dans une gamme limitée de
température et de vitesse de déplacement imposée. Dans ces conditions stationnaires, nous avons identifié deux
zones de dissipation d’énergie. La corrélation digitale d’image a permis de quantifier la dynamique de déformation
de l’intercalaire en aval du front et d’expliquer la grande quantité d’énergie dissipée. Enfin un modèle éléments
finis a été mis au point. Ce modèle a d’une part confirmé les observations expérimentales et d’autre part permit
d’explorer le voisinage du front de délamination.

Mots clés : Adhésion, Verre feuilleté, Rupture, Interface, Viscoélasticité, Grandes déformations